{
  "name": "LLM VRAM dataset",
  "version": "2026-09-25",
  "first_published": "2026-09-25",
  "updated": "2026-09-25",
  "description": "Estimated memory for 29 open-weight language models at six quantisations and each common context length, and how each fits 20 GPUs. Architecture values are read from each model's config.json; GPU specs are the manufacturers'.",
  "publisher": "Nodegrove",
  "homepage": "https://nodegrove.io/data",
  "repository": "https://github.com/nodegrove/llm-vram-dataset",
  "license": "CC-BY-4.0",
  "license_url": "https://creativecommons.org/licenses/by/4.0/",
  "doi": "10.5281/zenodo.22966137",
  "citation": "Nodegrove (2026). LLM VRAM dataset (version 2026-09-25) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.22966137",
  "findings": [
    {
      "headline": "Newer models pay far less for context.",
      "detail": "At 32k tokens, Qwen3 32B (2025, standard attention) spends 8.6 GB on its KV cache; its successor Qwen3.8 27B (2026, hybrid attention) spends 2.3 GB. Across all 29 models, each 1,000 tokens of context costs between 0.006 GB (Nemotron 3.5 Lightning 30B-A3B) and 0.328 GB (DeepSeek-R1 Distill Llama 70B)."
    },
    {
      "headline": "Active parameters set the speed, not the memory.",
      "detail": "Qwen3-Coder-Next 80B-A3B reads 3B parameters per token but keeps all 79.7B in memory: 48.9 GB at Q4_K_M with 8k context, against 22.4 GB for the dense Qwen3 32B."
    },
    {
      "headline": "24 of 29 models fit a 24 GB card at Q4_K_M with 8k context.",
      "detail": "The largest is Qwen3.6 35B-A3B at 22.4 GB, under the 22.8 GB line (95% of the card)."
    }
  ],
  "files": [
    {
      "name": "llm-vram-models.csv",
      "url": "https://nodegrove.io/data/llm-vram-models.csv",
      "rows": 29,
      "about": "One row per model: size, attention design, context window, licence and the source of every value"
    },
    {
      "name": "llm-vram-gpus.csv",
      "url": "https://nodegrove.io/data/llm-vram-gpus.csv",
      "rows": 20,
      "about": "One row per GPU: memory, the memory a runtime can use, bandwidth and the specification page"
    },
    {
      "name": "llm-vram-requirements.csv",
      "url": "https://nodegrove.io/data/llm-vram-requirements.csv",
      "rows": 1068,
      "about": "Memory per model, quantisation and context length, split into weights, KV cache and overhead"
    },
    {
      "name": "llm-vram-gpu-fit.csv",
      "url": "https://nodegrove.io/data/llm-vram-gpu-fit.csv",
      "rows": 3480,
      "about": "Every model on every GPU at every quantisation, at 8k context: verdict, longest context that fits and speed ceiling"
    },
    {
      "name": "llm-vram.json",
      "url": "https://nodegrove.io/data/llm-vram.json",
      "rows": null,
      "about": "Everything above in one file, with the method, its constants and each model's cache layout"
    }
  ],
  "method": {
    "total_gb": "weights_gb + kv_cache_gb + overhead_gb",
    "weights_gb": "params_b × bytes_per_param",
    "kv_cache_gb": "For each group of attention layers: layers × values cached per token × min(context, window) × 2 bytes (FP16) or 1 byte (Q8), plus fixed_state_gb. See kv_groups on each model",
    "overhead_gb": "0.5 + 0.04 × weights_gb",
    "fit": "A model fits when need_gb ≤ 0.95 × usable_gb; tight when need_gb > 0.85 × usable_gb",
    "tokens_per_second_ceiling": "0.7 × bandwidth_gb_s ÷ (active parameters, billions × bytes_per_param). Single stream, decode only; a ceiling, not a measurement",
    "constants": {
      "overhead_fixed_gb": 0.5,
      "overhead_share_of_weights": 0.04,
      "headroom": 0.95,
      "tight_above": 0.85,
      "apple_gpu_share": 0.75,
      "speed_efficiency": 0.7,
      "fit_context_tokens": 8192,
      "max_context_tokens": 131072
    }
  },
  "quants": [
    {
      "quant": "FP16",
      "bytes_per_param": 2,
      "note": "full precision, best quality, largest"
    },
    {
      "quant": "Q8_0",
      "bytes_per_param": 1.06,
      "note": "near-lossless"
    },
    {
      "quant": "Q6_K",
      "bytes_per_param": 0.82,
      "note": "very close to Q8"
    },
    {
      "quant": "Q5_K_M",
      "bytes_per_param": 0.71,
      "note": "good balance"
    },
    {
      "quant": "Q4_K_M",
      "bytes_per_param": 0.58,
      "note": "the common default; small quality loss"
    },
    {
      "quant": "Q3_K_M",
      "bytes_per_param": 0.47,
      "note": "noticeable loss; last resort"
    }
  ],
  "columns": {
    "models": {
      "model_id": "Stable identifier, also the page slug on nodegrove.io",
      "model": "Model name",
      "family": "Model family",
      "released": "Month the weights were published (YYYY-MM)",
      "params_b": "Total parameters, billions, as the checkpoint reports them. Mixture-of-experts models count every expert; multimodal checkpoints include the vision encoder",
      "active_params_b": "Parameters read per token, billions. Mixture-of-experts models only; blank for dense models",
      "attention": "How the model caches context: standard, sliding-window, hybrid or latent (see the method)",
      "layers": "Transformer layers (blocks)",
      "kv_heads": "Key/value heads of the main attention layers",
      "head_dim": "Dimension of each key/value head",
      "full_cache_layers": "Layers whose cache grows with the whole context",
      "window_layers": "Layers that only keep the last window_tokens tokens",
      "window_tokens": "Sliding-window length in tokens; blank if the model has none",
      "kv_cache_gb_per_1k_tokens": "Memory each extra 1,000 tokens of context costs, GB, with an FP16 cache once any windows are full",
      "fixed_state_gb": "Fixed recurrent state of linear-attention or Mamba layers, GB. Does not grow with context",
      "context_tokens": "Native context window in tokens (max_position_embeddings, or the model card where the two differ)",
      "license": "Licence of the weights, from the model card",
      "license_permissive": "true when the licence has no field-of-use or scale conditions",
      "role": "code or reasoning for specialist models; blank for general-purpose models",
      "successor_id": "Newer model in the same family and role, if there is one",
      "hf_repo": "Hugging Face repository the values were read from",
      "hf_gated": "true when config.json opens only after accepting the licence on Hugging Face",
      "config_url": "Link to the config.json the architecture values come from",
      "page_url": "The model page on nodegrove.io",
      "attention_detail": "The attention layout in one sentence",
      "notes": "Caveats that change how the numbers should be read"
    },
    "gpus": {
      "gpu_id": "Stable identifier, also the page slug on nodegrove.io",
      "gpu": "Card or machine",
      "kind": "consumer, workstation, datacenter, apple, or unified (non-Apple unified memory)",
      "maker": "NVIDIA, AMD, Apple or Intel",
      "memory_gb": "Memory from the manufacturer specification, GB (unified memory for Apple silicon)",
      "usable_gb": "Memory an inference runtime can address, GB. Apple silicon: about 75% of unified memory (observed; Apple publishes no figure). Other unified-memory machines: the share their maker documents, stated on the GPU page",
      "budget_gb": "95% of usable_gb: the line a model must fit under to count as fitting",
      "bandwidth_gb_s": "Memory bandwidth from the manufacturer specification, GB/s",
      "spec_url": "The manufacturer's specification page",
      "bandwidth_source_url": "Where the bandwidth figure is printed, when spec_url does not print it (a maker whitepaper, datasheet or announcement, or the data-rate source for a stated bus × rate calculation). Empty when spec_url prints it",
      "page_url": "The GPU page on nodegrove.io",
      "notes": "Caveats"
    },
    "requirements": {
      "model_id": "Joins llm-vram-models.csv",
      "model": "Model name",
      "quant": "FP16, Q8_0, Q6_K, Q5_K_M, Q4_K_M or Q3_K_M",
      "bytes_per_param": "Effective bytes per parameter at that quantisation",
      "context_tokens": "Tokens held in context",
      "weights_gb": "Parameters × bytes_per_param, GB",
      "kv_cache_gb": "KV cache stored at FP16, plus any fixed state, GB",
      "overhead_gb": "Runtime context and buffers: 0.5 GB + 4% of weights",
      "total_gb": "weights_gb + kv_cache_gb + overhead_gb",
      "total_gb_q8_cache": "The same total with the KV cache quantised to 8 bits"
    },
    "fit": {
      "model_id": "Joins llm-vram-models.csv",
      "model": "Model name",
      "gpu_id": "Joins llm-vram-gpus.csv",
      "gpu": "Card or machine",
      "quant": "Quantisation of the weights",
      "context_tokens": "Context the fit is computed at (8,192 tokens)",
      "need_gb": "Estimated memory at that quantisation and context, GB (total_gb in llm-vram-requirements.csv)",
      "verdict": "fits: need_gb is within the GPU's budget_gb. tight: it fits, but above 85% of usable memory. no: it does not fit",
      "max_context_tokens": "Longest context that fits at this quantisation with an FP16 cache, capped at 131,072 tokens or the model's window. 0 when the weights alone do not fit",
      "tokens_per_second_ceiling": "Upper bound on single-stream generation speed, tokens/s (see the method). Blank when the model cannot load"
    }
  },
  "models": [
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","family":"Llama","released":"2024-09","params_b":3.2,"active_params_b":null,"attention":"standard","layers":28,"kv_heads":8,"head_dim":128,"full_cache_layers":28,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.115,"fixed_state_gb":null,"context_tokens":131072,"license":"Llama Community License","license_permissive":false,"role":null,"successor_id":null,"hf_repo":"meta-llama/Llama-3.2-3B","hf_gated":true,"config_url":"https://huggingface.co/meta-llama/Llama-3.2-3B/blob/main/config.json","page_url":"https://nodegrove.io/models/llama-3.2-3b","attention_detail":"28 layers, all full attention, with 8 KV heads of dimension 128","notes":null,"kv_groups":[{"layers":28,"values_per_token":2048,"window_tokens":null}]},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","family":"Qwen","released":"2026-02","params_b":4.7,"active_params_b":null,"attention":"hybrid","layers":32,"kv_heads":4,"head_dim":256,"full_cache_layers":8,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.033,"fixed_state_gb":0.05,"context_tokens":262144,"license":"Apache 2.0","license_permissive":true,"role":null,"successor_id":null,"hf_repo":"Qwen/Qwen3.5-4B","hf_gated":false,"config_url":"https://huggingface.co/Qwen/Qwen3.5-4B/blob/main/config.json","page_url":"https://nodegrove.io/models/qwen3.5-4b","attention_detail":"32 layers, of which only 8 are full attention (4 KV heads of dimension 256) and keep a cache that grows with the conversation; the other 24 are Gated DeltaNet linear-attention layers with a small fixed state","notes":"The parameter count includes the vision encoder; a text-only GGUF is slightly smaller.","kv_groups":[{"layers":8,"values_per_token":2048,"window_tokens":null}]},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","family":"Llama","released":"2024-07","params_b":8,"active_params_b":null,"attention":"standard","layers":32,"kv_heads":8,"head_dim":128,"full_cache_layers":32,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.131,"fixed_state_gb":null,"context_tokens":131072,"license":"Llama Community License","license_permissive":false,"role":null,"successor_id":null,"hf_repo":"meta-llama/Llama-3.1-8B","hf_gated":true,"config_url":"https://huggingface.co/meta-llama/Llama-3.1-8B/blob/main/config.json","page_url":"https://nodegrove.io/models/llama-3.1-8b","attention_detail":"32 layers, all full attention, with 8 KV heads of dimension 128","notes":null,"kv_groups":[{"layers":32,"values_per_token":2048,"window_tokens":null}]},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","family":"Qwen","released":"2025-04","params_b":8.2,"active_params_b":null,"attention":"standard","layers":36,"kv_heads":8,"head_dim":128,"full_cache_layers":36,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.147,"fixed_state_gb":null,"context_tokens":40960,"license":"Apache 2.0","license_permissive":true,"role":null,"successor_id":"qwen3.5-9b","hf_repo":"Qwen/Qwen3-8B","hf_gated":false,"config_url":"https://huggingface.co/Qwen/Qwen3-8B/blob/main/config.json","page_url":"https://nodegrove.io/models/qwen3-8b","attention_detail":"36 layers, all full attention, with 8 KV heads of dimension 128","notes":null,"kv_groups":[{"layers":36,"values_per_token":2048,"window_tokens":null}]},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","family":"Qwen","released":"2026-02","params_b":9.7,"active_params_b":null,"attention":"hybrid","layers":32,"kv_heads":4,"head_dim":256,"full_cache_layers":8,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.033,"fixed_state_gb":0.05,"context_tokens":262144,"license":"Apache 2.0","license_permissive":true,"role":null,"successor_id":null,"hf_repo":"Qwen/Qwen3.5-9B","hf_gated":false,"config_url":"https://huggingface.co/Qwen/Qwen3.5-9B/blob/main/config.json","page_url":"https://nodegrove.io/models/qwen3.5-9b","attention_detail":"32 layers, of which only 8 are full attention (4 KV heads of dimension 256) and keep a cache that grows with the conversation; the other 24 are Gated DeltaNet linear-attention layers with a small fixed state","notes":"The parameter count includes the vision encoder; a text-only GGUF is slightly smaller.","kv_groups":[{"layers":8,"values_per_token":2048,"window_tokens":null}]},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","family":"Gemma","released":"2026-05","params_b":12,"active_params_b":null,"attention":"sliding-window","layers":48,"kv_heads":8,"head_dim":256,"full_cache_layers":8,"window_layers":40,"window_tokens":1024,"kv_cache_gb_per_1k_tokens":0.016,"fixed_state_gb":null,"context_tokens":262144,"license":"Apache 2.0","license_permissive":true,"role":null,"successor_id":null,"hf_repo":"google/gemma-4-12B-it","hf_gated":false,"config_url":"https://huggingface.co/google/gemma-4-12B-it/blob/main/config.json","page_url":"https://nodegrove.io/models/gemma-4-12b","attention_detail":"48 layers: 40 use sliding-window attention and only remember the last 1,024 tokens (8 KV heads of dimension 256), and 8 are global layers with a single KV head of dimension 512","notes":"A runtime without a sliding-window cache stores every layer at full length and needs far more than this at long context.","kv_groups":[{"layers":8,"values_per_token":1024,"window_tokens":null},{"layers":40,"values_per_token":4096,"window_tokens":1024}]},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","family":"Gemma","released":"2025-03","params_b":12.2,"active_params_b":null,"attention":"sliding-window","layers":48,"kv_heads":8,"head_dim":256,"full_cache_layers":8,"window_layers":40,"window_tokens":1024,"kv_cache_gb_per_1k_tokens":0.066,"fixed_state_gb":null,"context_tokens":131072,"license":"Gemma Terms of Use","license_permissive":false,"role":null,"successor_id":"gemma-4-12b","hf_repo":"google/gemma-3-12b-it","hf_gated":true,"config_url":"https://huggingface.co/google/gemma-3-12b-it/blob/main/config.json","page_url":"https://nodegrove.io/models/gemma-3-12b","attention_detail":"48 layers with 8 KV heads of dimension 256; five in every six use sliding-window attention and only remember the last 1,024 tokens, so 8 global layers carry the long context","notes":"A runtime without a sliding-window cache stores every layer at full length and needs far more than this at long context.","kv_groups":[{"layers":8,"values_per_token":4096,"window_tokens":null},{"layers":40,"values_per_token":4096,"window_tokens":1024}]},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","family":"Mistral","released":"2025-12","params_b":14,"active_params_b":null,"attention":"standard","layers":40,"kv_heads":8,"head_dim":128,"full_cache_layers":40,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.164,"fixed_state_gb":null,"context_tokens":262144,"license":"Apache 2.0","license_permissive":true,"role":null,"successor_id":null,"hf_repo":"mistralai/Ministral-3-14B-Instruct-2512","hf_gated":false,"config_url":"https://huggingface.co/mistralai/Ministral-3-14B-Instruct-2512/blob/main/config.json","page_url":"https://nodegrove.io/models/ministral-3-14b","attention_detail":"40 layers, all full attention, with 8 KV heads of dimension 128","notes":"The parameter count includes the vision encoder; a text-only GGUF is slightly smaller.","kv_groups":[{"layers":40,"values_per_token":2048,"window_tokens":null}]},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","family":"Qwen","released":"2025-04","params_b":14.8,"active_params_b":null,"attention":"standard","layers":40,"kv_heads":8,"head_dim":128,"full_cache_layers":40,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.164,"fixed_state_gb":null,"context_tokens":40960,"license":"Apache 2.0","license_permissive":true,"role":null,"successor_id":null,"hf_repo":"Qwen/Qwen3-14B","hf_gated":false,"config_url":"https://huggingface.co/Qwen/Qwen3-14B/blob/main/config.json","page_url":"https://nodegrove.io/models/qwen3-14b","attention_detail":"40 layers, all full attention, with 8 KV heads of dimension 128","notes":null,"kv_groups":[{"layers":40,"values_per_token":2048,"window_tokens":null}]},
    {"model_id":"phi-4","model":"Phi-4 14B","family":"Phi","released":"2024-12","params_b":14.7,"active_params_b":null,"attention":"standard","layers":40,"kv_heads":10,"head_dim":128,"full_cache_layers":40,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.205,"fixed_state_gb":null,"context_tokens":16384,"license":"MIT","license_permissive":true,"role":null,"successor_id":null,"hf_repo":"microsoft/phi-4","hf_gated":false,"config_url":"https://huggingface.co/microsoft/phi-4/blob/main/config.json","page_url":"https://nodegrove.io/models/phi-4","attention_detail":"40 layers, all full attention, with 10 KV heads of dimension 128","notes":null,"kv_groups":[{"layers":40,"values_per_token":2560,"window_tokens":null}]},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","family":"gpt-oss","released":"2025-08","params_b":21,"active_params_b":3.6,"attention":"sliding-window","layers":24,"kv_heads":8,"head_dim":64,"full_cache_layers":12,"window_layers":12,"window_tokens":128,"kv_cache_gb_per_1k_tokens":0.025,"fixed_state_gb":null,"context_tokens":131072,"license":"Apache 2.0","license_permissive":true,"role":null,"successor_id":null,"hf_repo":"openai/gpt-oss-20b","hf_gated":false,"config_url":"https://huggingface.co/openai/gpt-oss-20b/blob/main/config.json","page_url":"https://nodegrove.io/models/gpt-oss-20b","attention_detail":"24 layers with 8 KV heads of dimension 64, alternating full attention with a 128-token sliding window, so only half the layers keep a cache that grows","notes":"Ships in MXFP4; roughly 4.25 bits per expert weight. Choose the 4-bit row.","kv_groups":[{"layers":12,"values_per_token":1024,"window_tokens":null},{"layers":12,"values_per_token":1024,"window_tokens":128}]},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","family":"Mistral","released":"2025-12","params_b":24,"active_params_b":null,"attention":"standard","layers":40,"kv_heads":8,"head_dim":128,"full_cache_layers":40,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.164,"fixed_state_gb":null,"context_tokens":262144,"license":"Apache 2.0","license_permissive":true,"role":"code","successor_id":null,"hf_repo":"mistralai/Devstral-Small-2-24B-Instruct-2512","hf_gated":false,"config_url":"https://huggingface.co/mistralai/Devstral-Small-2-24B-Instruct-2512/blob/main/config.json","page_url":"https://nodegrove.io/models/devstral-small-2-24b","attention_detail":"40 layers, all full attention, with 8 KV heads of dimension 128","notes":"Mistral's coding model. The parameter count includes the vision encoder; a text-only GGUF is slightly smaller.","kv_groups":[{"layers":40,"values_per_token":2048,"window_tokens":null}]},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","family":"Mistral","released":"2025-03","params_b":24,"active_params_b":null,"attention":"standard","layers":40,"kv_heads":8,"head_dim":128,"full_cache_layers":40,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.164,"fixed_state_gb":null,"context_tokens":131072,"license":"Apache 2.0","license_permissive":true,"role":null,"successor_id":"mistral-small-4-119b","hf_repo":"mistralai/Mistral-Small-3.1-24B-Instruct-2503","hf_gated":false,"config_url":"https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/blob/main/config.json","page_url":"https://nodegrove.io/models/mistral-small-3.1-24b","attention_detail":"40 layers, all full attention, with 8 KV heads of dimension 128","notes":null,"kv_groups":[{"layers":40,"values_per_token":2048,"window_tokens":null}]},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","family":"Gemma","released":"2026-03","params_b":25.8,"active_params_b":3.8,"attention":"sliding-window","layers":30,"kv_heads":8,"head_dim":256,"full_cache_layers":5,"window_layers":25,"window_tokens":1024,"kv_cache_gb_per_1k_tokens":0.02,"fixed_state_gb":null,"context_tokens":262144,"license":"Apache 2.0","license_permissive":true,"role":null,"successor_id":null,"hf_repo":"google/gemma-4-26B-A4B-it","hf_gated":false,"config_url":"https://huggingface.co/google/gemma-4-26B-A4B-it/blob/main/config.json","page_url":"https://nodegrove.io/models/gemma-4-26b-a4b","attention_detail":"30 layers: 25 use sliding-window attention over the last 1,024 tokens (8 KV heads of dimension 256), and 5 are global layers with 2 KV heads of dimension 512","notes":"All 128 experts stay in memory; 3.8B are active per token. The parameter count includes the vision encoder; a text-only GGUF is slightly smaller.","kv_groups":[{"layers":5,"values_per_token":2048,"window_tokens":null},{"layers":25,"values_per_token":4096,"window_tokens":1024}]},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","family":"Gemma","released":"2025-03","params_b":27.4,"active_params_b":null,"attention":"sliding-window","layers":62,"kv_heads":16,"head_dim":128,"full_cache_layers":10,"window_layers":52,"window_tokens":1024,"kv_cache_gb_per_1k_tokens":0.082,"fixed_state_gb":null,"context_tokens":131072,"license":"Gemma Terms of Use","license_permissive":false,"role":null,"successor_id":"gemma-4-31b","hf_repo":"google/gemma-3-27b-it","hf_gated":true,"config_url":"https://huggingface.co/google/gemma-3-27b-it/blob/main/config.json","page_url":"https://nodegrove.io/models/gemma-3-27b","attention_detail":"62 layers with 16 KV heads of dimension 128; five in every six use sliding-window attention and only remember the last 1,024 tokens, so 10 global layers carry the long context","notes":"A runtime without a sliding-window cache stores every layer at full length and needs far more than this at long context.","kv_groups":[{"layers":10,"values_per_token":4096,"window_tokens":null},{"layers":52,"values_per_token":4096,"window_tokens":1024}]},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","family":"Qwen","released":"2026-08","params_b":27.8,"active_params_b":null,"attention":"hybrid","layers":64,"kv_heads":4,"head_dim":256,"full_cache_layers":16,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.066,"fixed_state_gb":0.15,"context_tokens":262144,"license":"Apache 2.0","license_permissive":true,"role":null,"successor_id":null,"hf_repo":"Qwen/Qwen3.8-27B","hf_gated":false,"config_url":"https://huggingface.co/Qwen/Qwen3.8-27B/blob/main/config.json","page_url":"https://nodegrove.io/models/qwen3.8-27b","attention_detail":"64 layers, of which only 16 are full attention (4 KV heads of dimension 256) and keep a cache that grows with the conversation; the other 48 are Gated DeltaNet linear-attention layers with a fixed state of about 0.15 GB","notes":"The parameter count includes the vision encoder; a text-only GGUF is slightly smaller.","kv_groups":[{"layers":16,"values_per_token":2048,"window_tokens":null}]},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","family":"Qwen","released":"2025-04","params_b":30.5,"active_params_b":3.3,"attention":"standard","layers":48,"kv_heads":4,"head_dim":128,"full_cache_layers":48,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.098,"fixed_state_gb":null,"context_tokens":40960,"license":"Apache 2.0","license_permissive":true,"role":null,"successor_id":"qwen3.6-35b-a3b","hf_repo":"Qwen/Qwen3-30B-A3B","hf_gated":false,"config_url":"https://huggingface.co/Qwen/Qwen3-30B-A3B/blob/main/config.json","page_url":"https://nodegrove.io/models/qwen3-30b-a3b","attention_detail":"48 layers, all full attention, with 4 KV heads of dimension 128","notes":"All experts stay in memory; only 3.3B are active per token, so it is fast for its size.","kv_groups":[{"layers":48,"values_per_token":1024,"window_tokens":null}]},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","family":"GLM","released":"2026-01","params_b":31.2,"active_params_b":3,"attention":"latent","layers":47,"kv_heads":20,"head_dim":256,"full_cache_layers":47,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.054,"fixed_state_gb":null,"context_tokens":202752,"license":"MIT","license_permissive":true,"role":null,"successor_id":null,"hf_repo":"zai-org/GLM-4.7-Flash","hf_gated":false,"config_url":"https://huggingface.co/zai-org/GLM-4.7-Flash/blob/main/config.json","page_url":"https://nodegrove.io/models/glm-4.7-flash","attention_detail":"47 layers using multi-head latent attention: instead of keys and values for each of its 20 heads, every layer caches one compressed vector of 576 values per token","notes":"A runtime that does not implement the latent cache stores full keys and values and needs several times more context memory.","kv_groups":[{"layers":47,"values_per_token":576,"window_tokens":null}]},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","family":"Gemma","released":"2026-03","params_b":31.3,"active_params_b":null,"attention":"sliding-window","layers":60,"kv_heads":16,"head_dim":256,"full_cache_layers":10,"window_layers":50,"window_tokens":1024,"kv_cache_gb_per_1k_tokens":0.082,"fixed_state_gb":null,"context_tokens":262144,"license":"Apache 2.0","license_permissive":true,"role":null,"successor_id":null,"hf_repo":"google/gemma-4-31B-it","hf_gated":false,"config_url":"https://huggingface.co/google/gemma-4-31B-it/blob/main/config.json","page_url":"https://nodegrove.io/models/gemma-4-31b","attention_detail":"60 layers: 50 use sliding-window attention over the last 1,024 tokens (16 KV heads of dimension 256), and 10 are global layers with 4 KV heads of dimension 512","notes":"A runtime without a sliding-window cache stores every layer at full length and needs far more than this at long context. The parameter count includes the vision encoder; a text-only GGUF is slightly smaller.","kv_groups":[{"layers":10,"values_per_token":4096,"window_tokens":null},{"layers":50,"values_per_token":8192,"window_tokens":1024}]},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","family":"Nemotron","released":"2026-08","params_b":31.6,"active_params_b":3,"attention":"hybrid","layers":52,"kv_heads":2,"head_dim":128,"full_cache_layers":6,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.006,"fixed_state_gb":0.05,"context_tokens":262144,"license":"OpenMDW 1.1","license_permissive":false,"role":null,"successor_id":null,"hf_repo":"nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16","hf_gated":false,"config_url":"https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/main/config.json","page_url":"https://nodegrove.io/models/nemotron-3.5-lightning-30b-a3b","attention_detail":"52 blocks, of which only 6 are attention layers (2 KV heads of dimension 128); 23 are Mamba-2 layers with a small fixed state and 23 are mixture-of-experts layers with no cache at all","notes":null,"kv_groups":[{"layers":6,"values_per_token":512,"window_tokens":null}]},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","family":"Qwen","released":"2025-04","params_b":32.8,"active_params_b":null,"attention":"standard","layers":64,"kv_heads":8,"head_dim":128,"full_cache_layers":64,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.262,"fixed_state_gb":null,"context_tokens":40960,"license":"Apache 2.0","license_permissive":true,"role":null,"successor_id":"qwen3.8-27b","hf_repo":"Qwen/Qwen3-32B","hf_gated":false,"config_url":"https://huggingface.co/Qwen/Qwen3-32B/blob/main/config.json","page_url":"https://nodegrove.io/models/qwen3-32b","attention_detail":"64 layers, all full attention, with 8 KV heads of dimension 128","notes":null,"kv_groups":[{"layers":64,"values_per_token":2048,"window_tokens":null}]},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","family":"Qwen","released":"2024-11","params_b":32.8,"active_params_b":null,"attention":"standard","layers":64,"kv_heads":8,"head_dim":128,"full_cache_layers":64,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.262,"fixed_state_gb":null,"context_tokens":32768,"license":"Apache 2.0","license_permissive":true,"role":"code","successor_id":"qwen3-coder-next","hf_repo":"Qwen/Qwen2.5-Coder-32B-Instruct","hf_gated":false,"config_url":"https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct/blob/main/config.json","page_url":"https://nodegrove.io/models/qwen2.5-coder-32b","attention_detail":"64 layers, all full attention, with 8 KV heads of dimension 128","notes":null,"kv_groups":[{"layers":64,"values_per_token":2048,"window_tokens":null}]},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","family":"DeepSeek","released":"2025-01","params_b":32.8,"active_params_b":null,"attention":"standard","layers":64,"kv_heads":8,"head_dim":128,"full_cache_layers":64,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.262,"fixed_state_gb":null,"context_tokens":131072,"license":"MIT","license_permissive":true,"role":"reasoning","successor_id":null,"hf_repo":"deepseek-ai/DeepSeek-R1-Distill-Qwen-32B","hf_gated":false,"config_url":"https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-32B/blob/main/config.json","page_url":"https://nodegrove.io/models/deepseek-r1-distill-qwen-32b","attention_detail":"64 layers, all full attention, with 8 KV heads of dimension 128","notes":null,"kv_groups":[{"layers":64,"values_per_token":2048,"window_tokens":null}]},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","family":"Qwen","released":"2026-04","params_b":36,"active_params_b":3,"attention":"hybrid","layers":40,"kv_heads":2,"head_dim":256,"full_cache_layers":10,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.02,"fixed_state_gb":0.06,"context_tokens":262144,"license":"Apache 2.0","license_permissive":true,"role":null,"successor_id":null,"hf_repo":"Qwen/Qwen3.6-35B-A3B","hf_gated":false,"config_url":"https://huggingface.co/Qwen/Qwen3.6-35B-A3B/blob/main/config.json","page_url":"https://nodegrove.io/models/qwen3.6-35b-a3b","attention_detail":"40 layers, of which only 10 are full attention (2 KV heads of dimension 256) and keep a cache that grows with the conversation; the other 30 are Gated DeltaNet linear-attention layers with a small fixed state","notes":"All 256 experts stay in memory; 3B are active per token. The parameter count includes the vision encoder; a text-only GGUF is slightly smaller.","kv_groups":[{"layers":10,"values_per_token":1024,"window_tokens":null}]},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","family":"Llama","released":"2024-12","params_b":70.6,"active_params_b":null,"attention":"standard","layers":80,"kv_heads":8,"head_dim":128,"full_cache_layers":80,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.328,"fixed_state_gb":null,"context_tokens":131072,"license":"Llama Community License","license_permissive":false,"role":null,"successor_id":null,"hf_repo":"meta-llama/Llama-3.3-70B-Instruct","hf_gated":true,"config_url":"https://huggingface.co/meta-llama/Llama-3.3-70B-Instruct/blob/main/config.json","page_url":"https://nodegrove.io/models/llama-3.3-70b","attention_detail":"80 layers, all full attention, with 8 KV heads of dimension 128","notes":null,"kv_groups":[{"layers":80,"values_per_token":2048,"window_tokens":null}]},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","family":"DeepSeek","released":"2025-01","params_b":70.6,"active_params_b":null,"attention":"standard","layers":80,"kv_heads":8,"head_dim":128,"full_cache_layers":80,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.328,"fixed_state_gb":null,"context_tokens":131072,"license":"MIT","license_permissive":true,"role":"reasoning","successor_id":null,"hf_repo":"deepseek-ai/DeepSeek-R1-Distill-Llama-70B","hf_gated":false,"config_url":"https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Llama-70B/blob/main/config.json","page_url":"https://nodegrove.io/models/deepseek-r1-distill-llama-70b","attention_detail":"80 layers, all full attention, with 8 KV heads of dimension 128","notes":null,"kv_groups":[{"layers":80,"values_per_token":2048,"window_tokens":null}]},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","family":"Qwen","released":"2026-01","params_b":79.7,"active_params_b":3,"attention":"hybrid","layers":48,"kv_heads":2,"head_dim":256,"full_cache_layers":12,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.025,"fixed_state_gb":0.08,"context_tokens":262144,"license":"Apache 2.0","license_permissive":true,"role":"code","successor_id":null,"hf_repo":"Qwen/Qwen3-Coder-Next","hf_gated":false,"config_url":"https://huggingface.co/Qwen/Qwen3-Coder-Next/blob/main/config.json","page_url":"https://nodegrove.io/models/qwen3-coder-next","attention_detail":"48 layers, of which only 12 are full attention (2 KV heads of dimension 256) and keep a cache that grows with the conversation; the other 36 are Gated DeltaNet linear-attention layers with a small fixed state","notes":"All 512 experts stay in memory; 3B are active per token, so it answers like a small model once it is loaded.","kv_groups":[{"layers":12,"values_per_token":1024,"window_tokens":null}]},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","family":"gpt-oss","released":"2025-08","params_b":117,"active_params_b":5.1,"attention":"sliding-window","layers":36,"kv_heads":8,"head_dim":64,"full_cache_layers":18,"window_layers":18,"window_tokens":128,"kv_cache_gb_per_1k_tokens":0.037,"fixed_state_gb":null,"context_tokens":131072,"license":"Apache 2.0","license_permissive":true,"role":null,"successor_id":null,"hf_repo":"openai/gpt-oss-120b","hf_gated":false,"config_url":"https://huggingface.co/openai/gpt-oss-120b/blob/main/config.json","page_url":"https://nodegrove.io/models/gpt-oss-120b","attention_detail":"36 layers with 8 KV heads of dimension 64, alternating full attention with a 128-token sliding window, so only half the layers keep a cache that grows","notes":"Ships in MXFP4; OpenAI states it runs on a single 80 GB GPU. Choose the 4-bit row.","kv_groups":[{"layers":18,"values_per_token":1024,"window_tokens":null},{"layers":18,"values_per_token":1024,"window_tokens":128}]},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","family":"Mistral","released":"2026-03","params_b":119,"active_params_b":6.5,"attention":"latent","layers":36,"kv_heads":32,"head_dim":128,"full_cache_layers":36,"window_layers":0,"window_tokens":null,"kv_cache_gb_per_1k_tokens":0.023,"fixed_state_gb":null,"context_tokens":262144,"license":"Apache 2.0","license_permissive":true,"role":null,"successor_id":null,"hf_repo":"mistralai/Mistral-Small-4-119B-2603","hf_gated":false,"config_url":"https://huggingface.co/mistralai/Mistral-Small-4-119B-2603/blob/main/config.json","page_url":"https://nodegrove.io/models/mistral-small-4-119b","attention_detail":"36 layers using multi-head latent attention: instead of keys and values for each of its 32 heads, every layer caches one compressed vector of 320 values per token","notes":"Small in name only: all 128 experts stay in memory. A runtime that does not implement the latent cache needs several times more context memory. The parameter count includes the vision encoder; a text-only GGUF is slightly smaller.","kv_groups":[{"layers":36,"values_per_token":320,"window_tokens":null}]}
  ],
  "gpus": [
    {"gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","kind":"consumer","maker":"NVIDIA","memory_gb":12,"usable_gb":12,"budget_gb":11.4,"bandwidth_gb_s":360,"spec_url":"https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3060-3060ti/","bandwidth_source_url":"https://www.nvidia.com/en-us/geforce/news/game-on-you-asked-we-answered-qa/","page_url":"https://nodegrove.io/gpus/rtx-3060-12","notes":null},
    {"gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","kind":"consumer","maker":"NVIDIA","memory_gb":16,"usable_gb":16,"budget_gb":15.2,"bandwidth_gb_s":288,"spec_url":"https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4060-4060ti/","bandwidth_source_url":"https://marketplace.nvidia.com/en-us/consumer/graphics-cards/msi-geforce-rtx-4060-ti-gaming-x-slim-16g-graphics-card-nvidia-rtx-4060-ti-16gb-gddr6-memory-18gbps-pcie-4-0-twin-frozr-9-rgb-dlss3/","page_url":"https://nodegrove.io/gpus/rtx-4060-ti-16","notes":null},
    {"gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","kind":"consumer","maker":"NVIDIA","memory_gb":12,"usable_gb":12,"budget_gb":11.4,"bandwidth_gb_s":504,"spec_url":"https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4070-family/","bandwidth_source_url":"https://www.nvidia.com/en-us/geforce/news/rtx-50-series-graphics-cards-gpu-laptop-announcements/","page_url":"https://nodegrove.io/gpus/rtx-4070","notes":null},
    {"gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","kind":"consumer","maker":"NVIDIA","memory_gb":24,"usable_gb":24,"budget_gb":22.8,"bandwidth_gb_s":936,"spec_url":"https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3090-3090ti/","bandwidth_source_url":"https://www.nvidia.com/content/PDF/nvidia-ampere-ga-102-gpu-architecture-whitepaper-v2.pdf","page_url":"https://nodegrove.io/gpus/rtx-3090","notes":null},
    {"gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","kind":"consumer","maker":"NVIDIA","memory_gb":24,"usable_gb":24,"budget_gb":22.8,"bandwidth_gb_s":1008,"spec_url":"https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4090/","bandwidth_source_url":"https://images.nvidia.com/aem-dam/Solutions/geforce/ada/nvidia-ada-gpu-architecture.pdf","page_url":"https://nodegrove.io/gpus/rtx-4090","notes":null},
    {"gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","kind":"consumer","maker":"NVIDIA","memory_gb":32,"usable_gb":32,"budget_gb":30.4,"bandwidth_gb_s":1792,"spec_url":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5090/","bandwidth_source_url":"https://www.nvidia.com/en-us/geforce/graphics-cards/compare/","page_url":"https://nodegrove.io/gpus/rtx-5090","notes":null},
    {"gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","kind":"workstation","maker":"NVIDIA","memory_gb":48,"usable_gb":48,"budget_gb":45.6,"bandwidth_gb_s":960,"spec_url":"https://www.nvidia.com/en-us/design-visualization/rtx-6000/","bandwidth_source_url":"https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/rtx-6000/proviz-print-rtx6000-datasheet-web-2504660.pdf","page_url":"https://nodegrove.io/gpus/rtx-6000-ada","notes":null},
    {"gpu_id":"l40s","gpu":"L40S 48 GB","kind":"datacenter","maker":"NVIDIA","memory_gb":48,"usable_gb":48,"budget_gb":45.6,"bandwidth_gb_s":864,"spec_url":"https://www.nvidia.com/en-us/data-center/l40s/","bandwidth_source_url":null,"page_url":"https://nodegrove.io/gpus/l40s","notes":null},
    {"gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","kind":"workstation","maker":"NVIDIA","memory_gb":96,"usable_gb":96,"budget_gb":91.2,"bandwidth_gb_s":1792,"spec_url":"https://www.nvidia.com/en-us/products/workstations/professional-desktop-gpus/rtx-pro-6000/","bandwidth_source_url":null,"page_url":"https://nodegrove.io/gpus/rtx-pro-6000","notes":null},
    {"gpu_id":"a100-80","gpu":"A100 80 GB","kind":"datacenter","maker":"NVIDIA","memory_gb":80,"usable_gb":80,"budget_gb":76,"bandwidth_gb_s":2039,"spec_url":"https://www.nvidia.com/en-us/data-center/a100/","bandwidth_source_url":null,"page_url":"https://nodegrove.io/gpus/a100-80","notes":null},
    {"gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","kind":"datacenter","maker":"NVIDIA","memory_gb":80,"usable_gb":80,"budget_gb":76,"bandwidth_gb_s":3350,"spec_url":"https://www.nvidia.com/en-us/data-center/h100/","bandwidth_source_url":null,"page_url":"https://nodegrove.io/gpus/h100-sxm","notes":null},
    {"gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","kind":"apple","maker":"Apple","memory_gb":192,"usable_gb":144,"budget_gb":136.8,"bandwidth_gb_s":800,"spec_url":"https://support.apple.com/en-us/111835","bandwidth_source_url":null,"page_url":"https://nodegrove.io/gpus/m2-ultra","notes":"Unified memory: the model shares RAM with the system. Up to 192 GB configurable; prompt processing is slower than on NVIDIA."},
    {"gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","kind":"apple","maker":"Apple","memory_gb":128,"usable_gb":96,"budget_gb":91.2,"bandwidth_gb_s":546,"spec_url":"https://support.apple.com/en-us/121554","bandwidth_source_url":null,"page_url":"https://nodegrove.io/gpus/m4-max","notes":"Unified memory, up to 128 GB configurable."},
    {"gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","kind":"consumer","maker":"NVIDIA","memory_gb":16,"usable_gb":16,"budget_gb":15.2,"bandwidth_gb_s":448,"spec_url":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5060-family/","bandwidth_source_url":"https://www.nvidia.com/en-us/geforce/graphics-cards/compare/","page_url":"https://nodegrove.io/gpus/rtx-5060-ti-16","notes":null},
    {"gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","kind":"consumer","maker":"AMD","memory_gb":24,"usable_gb":24,"budget_gb":22.8,"bandwidth_gb_s":960,"spec_url":"https://www.amd.com/en/products/graphics/desktops/radeon/7000-series/amd-radeon-rx-7900xtx.html","bandwidth_source_url":null,"page_url":"https://nodegrove.io/gpus/rx-7900-xtx","notes":null},
    {"gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","kind":"unified","maker":"NVIDIA","memory_gb":128,"usable_gb":126,"budget_gb":119.7,"bandwidth_gb_s":273,"spec_url":"https://www.nvidia.com/en-us/products/workstations/dgx-spark/","bandwidth_source_url":null,"page_url":"https://nodegrove.io/gpus/dgx-spark","notes":"Unified memory shared with the Arm CPU and DGX OS. A 64 GB version is sold through partners from 23 October 2026."},
    {"gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","kind":"unified","maker":"AMD","memory_gb":128,"usable_gb":96,"budget_gb":91.2,"bandwidth_gb_s":256,"spec_url":"https://www.amd.com/en/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-max-plus-395.html","bandwidth_source_url":"https://www.amd.com/en/products/processors/desktops/ryzen/ryzen-ai-halo/ryzen-ai-max-plus-395.html","page_url":"https://nodegrove.io/gpus/ryzen-ai-max-395","notes":"Unified memory shared with the CPU. The power limit (45 to 120 W) is set by the laptop or mini-PC maker, so the same chip performs differently by system."},
    {"gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","kind":"apple","maker":"Apple","memory_gb":128,"usable_gb":96,"budget_gb":91.2,"bandwidth_gb_s":614,"spec_url":"https://www.apple.com/macbook-pro/specs/","bandwidth_source_url":null,"page_url":"https://nodegrove.io/gpus/m5-max","notes":"Unified memory. 128 GB and 614 GB/s need the 40-core GPU; the 32-core GPU version is 36 GB at 460 GB/s."},
    {"gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","kind":"apple","maker":"Apple","memory_gb":512,"usable_gb":384,"budget_gb":364.8,"bandwidth_gb_s":819,"spec_url":"https://www.apple.com/newsroom/2025/03/apple-unveils-new-mac-studio-the-most-powerful-mac-ever/","bandwidth_source_url":"https://support.apple.com/en-us/122211","page_url":"https://nodegrove.io/gpus/m3-ultra","notes":"Unified memory. Launched configurable to 512 GB (Apple newsroom, March 2025); Apple's current spec page lists up to 256 GB."},
    {"gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","kind":"apple","maker":"Apple","memory_gb":512,"usable_gb":384,"budget_gb":364.8,"bandwidth_gb_s":1200,"spec_url":"https://www.apple.com/mac-studio/specs/","bandwidth_source_url":null,"page_url":"https://nodegrove.io/gpus/m5-ultra","notes":"Unified memory. 512 GB needs the 80-core GPU chip and ships in late October 2026; otherwise 96 or 256 GB, at the same 1.2 TB/s."}
  ],
  "requirements": [
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":6.4,"kv_cache_gb":0.47,"overhead_gb":0.76,"total_gb":7.63,"total_gb_q8_cache":7.39},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":6.4,"kv_cache_gb":0.94,"overhead_gb":0.76,"total_gb":8.1,"total_gb_q8_cache":7.63},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":6.4,"kv_cache_gb":1.88,"overhead_gb":0.76,"total_gb":9.04,"total_gb_q8_cache":8.1},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":6.4,"kv_cache_gb":3.76,"overhead_gb":0.76,"total_gb":10.91,"total_gb_q8_cache":9.04},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":6.4,"kv_cache_gb":7.52,"overhead_gb":0.76,"total_gb":14.67,"total_gb_q8_cache":10.91},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":6.4,"kv_cache_gb":15.03,"overhead_gb":0.76,"total_gb":22.19,"total_gb_q8_cache":14.67},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":3.39,"kv_cache_gb":0.47,"overhead_gb":0.64,"total_gb":4.5,"total_gb_q8_cache":4.26},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":3.39,"kv_cache_gb":0.94,"overhead_gb":0.64,"total_gb":4.97,"total_gb_q8_cache":4.5},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":3.39,"kv_cache_gb":1.88,"overhead_gb":0.64,"total_gb":5.91,"total_gb_q8_cache":4.97},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":3.39,"kv_cache_gb":3.76,"overhead_gb":0.64,"total_gb":7.79,"total_gb_q8_cache":5.91},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":3.39,"kv_cache_gb":7.52,"overhead_gb":0.64,"total_gb":11.54,"total_gb_q8_cache":7.79},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":3.39,"kv_cache_gb":15.03,"overhead_gb":0.64,"total_gb":19.06,"total_gb_q8_cache":11.54},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":2.62,"kv_cache_gb":0.47,"overhead_gb":0.6,"total_gb":3.7,"total_gb_q8_cache":3.46},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":2.62,"kv_cache_gb":0.94,"overhead_gb":0.6,"total_gb":4.17,"total_gb_q8_cache":3.7},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":2.62,"kv_cache_gb":1.88,"overhead_gb":0.6,"total_gb":5.11,"total_gb_q8_cache":4.17},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":2.62,"kv_cache_gb":3.76,"overhead_gb":0.6,"total_gb":6.99,"total_gb_q8_cache":5.11},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":2.62,"kv_cache_gb":7.52,"overhead_gb":0.6,"total_gb":10.75,"total_gb_q8_cache":6.99},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":2.62,"kv_cache_gb":15.03,"overhead_gb":0.6,"total_gb":18.26,"total_gb_q8_cache":10.75},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":2.27,"kv_cache_gb":0.47,"overhead_gb":0.59,"total_gb":3.33,"total_gb_q8_cache":3.1},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":2.27,"kv_cache_gb":0.94,"overhead_gb":0.59,"total_gb":3.8,"total_gb_q8_cache":3.33},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":2.27,"kv_cache_gb":1.88,"overhead_gb":0.59,"total_gb":4.74,"total_gb_q8_cache":3.8},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":2.27,"kv_cache_gb":3.76,"overhead_gb":0.59,"total_gb":6.62,"total_gb_q8_cache":4.74},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":2.27,"kv_cache_gb":7.52,"overhead_gb":0.59,"total_gb":10.38,"total_gb_q8_cache":6.62},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":2.27,"kv_cache_gb":15.03,"overhead_gb":0.59,"total_gb":17.9,"total_gb_q8_cache":10.38},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":1.86,"kv_cache_gb":0.47,"overhead_gb":0.57,"total_gb":2.9,"total_gb_q8_cache":2.67},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":1.86,"kv_cache_gb":0.94,"overhead_gb":0.57,"total_gb":3.37,"total_gb_q8_cache":2.9},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":1.86,"kv_cache_gb":1.88,"overhead_gb":0.57,"total_gb":4.31,"total_gb_q8_cache":3.37},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":1.86,"kv_cache_gb":3.76,"overhead_gb":0.57,"total_gb":6.19,"total_gb_q8_cache":4.31},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":1.86,"kv_cache_gb":7.52,"overhead_gb":0.57,"total_gb":9.95,"total_gb_q8_cache":6.19},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":1.86,"kv_cache_gb":15.03,"overhead_gb":0.57,"total_gb":17.46,"total_gb_q8_cache":9.95},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":1.5,"kv_cache_gb":0.47,"overhead_gb":0.56,"total_gb":2.53,"total_gb_q8_cache":2.3},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":1.5,"kv_cache_gb":0.94,"overhead_gb":0.56,"total_gb":3,"total_gb_q8_cache":2.53},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":1.5,"kv_cache_gb":1.88,"overhead_gb":0.56,"total_gb":3.94,"total_gb_q8_cache":3},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":1.5,"kv_cache_gb":3.76,"overhead_gb":0.56,"total_gb":5.82,"total_gb_q8_cache":3.94},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":1.5,"kv_cache_gb":7.52,"overhead_gb":0.56,"total_gb":9.58,"total_gb_q8_cache":5.82},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":1.5,"kv_cache_gb":15.03,"overhead_gb":0.56,"total_gb":17.1,"total_gb_q8_cache":9.58},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":9.4,"kv_cache_gb":0.18,"overhead_gb":0.88,"total_gb":10.46,"total_gb_q8_cache":10.39},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":9.4,"kv_cache_gb":0.32,"overhead_gb":0.88,"total_gb":10.59,"total_gb_q8_cache":10.46},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":9.4,"kv_cache_gb":0.59,"overhead_gb":0.88,"total_gb":10.86,"total_gb_q8_cache":10.59},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":9.4,"kv_cache_gb":1.12,"overhead_gb":0.88,"total_gb":11.4,"total_gb_q8_cache":10.86},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":9.4,"kv_cache_gb":2.2,"overhead_gb":0.88,"total_gb":12.47,"total_gb_q8_cache":11.4},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":9.4,"kv_cache_gb":4.34,"overhead_gb":0.88,"total_gb":14.62,"total_gb_q8_cache":12.47},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"FP16","bytes_per_param":2,"context_tokens":262144,"weights_gb":9.4,"kv_cache_gb":8.64,"overhead_gb":0.88,"total_gb":18.92,"total_gb_q8_cache":14.62},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":4.98,"kv_cache_gb":0.18,"overhead_gb":0.7,"total_gb":5.87,"total_gb_q8_cache":5.8},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":4.98,"kv_cache_gb":0.32,"overhead_gb":0.7,"total_gb":6,"total_gb_q8_cache":5.87},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":4.98,"kv_cache_gb":0.59,"overhead_gb":0.7,"total_gb":6.27,"total_gb_q8_cache":6},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":4.98,"kv_cache_gb":1.12,"overhead_gb":0.7,"total_gb":6.81,"total_gb_q8_cache":6.27},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":4.98,"kv_cache_gb":2.2,"overhead_gb":0.7,"total_gb":7.88,"total_gb_q8_cache":6.81},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":4.98,"kv_cache_gb":4.34,"overhead_gb":0.7,"total_gb":10.03,"total_gb_q8_cache":7.88},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":262144,"weights_gb":4.98,"kv_cache_gb":8.64,"overhead_gb":0.7,"total_gb":14.32,"total_gb_q8_cache":10.03},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":3.85,"kv_cache_gb":0.18,"overhead_gb":0.65,"total_gb":4.69,"total_gb_q8_cache":4.63},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":3.85,"kv_cache_gb":0.32,"overhead_gb":0.65,"total_gb":4.83,"total_gb_q8_cache":4.69},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":3.85,"kv_cache_gb":0.59,"overhead_gb":0.65,"total_gb":5.1,"total_gb_q8_cache":4.83},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":3.85,"kv_cache_gb":1.12,"overhead_gb":0.65,"total_gb":5.63,"total_gb_q8_cache":5.1},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":3.85,"kv_cache_gb":2.2,"overhead_gb":0.65,"total_gb":6.71,"total_gb_q8_cache":5.63},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":3.85,"kv_cache_gb":4.34,"overhead_gb":0.65,"total_gb":8.85,"total_gb_q8_cache":6.71},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":262144,"weights_gb":3.85,"kv_cache_gb":8.64,"overhead_gb":0.65,"total_gb":13.15,"total_gb_q8_cache":8.85},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":3.34,"kv_cache_gb":0.18,"overhead_gb":0.63,"total_gb":4.15,"total_gb_q8_cache":4.09},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":3.34,"kv_cache_gb":0.32,"overhead_gb":0.63,"total_gb":4.29,"total_gb_q8_cache":4.15},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":3.34,"kv_cache_gb":0.59,"overhead_gb":0.63,"total_gb":4.56,"total_gb_q8_cache":4.29},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":3.34,"kv_cache_gb":1.12,"overhead_gb":0.63,"total_gb":5.09,"total_gb_q8_cache":4.56},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":3.34,"kv_cache_gb":2.2,"overhead_gb":0.63,"total_gb":6.17,"total_gb_q8_cache":5.09},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":3.34,"kv_cache_gb":4.34,"overhead_gb":0.63,"total_gb":8.32,"total_gb_q8_cache":6.17},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":262144,"weights_gb":3.34,"kv_cache_gb":8.64,"overhead_gb":0.63,"total_gb":12.61,"total_gb_q8_cache":8.32},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":2.73,"kv_cache_gb":0.18,"overhead_gb":0.61,"total_gb":3.52,"total_gb_q8_cache":3.45},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":2.73,"kv_cache_gb":0.32,"overhead_gb":0.61,"total_gb":3.65,"total_gb_q8_cache":3.52},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":2.73,"kv_cache_gb":0.59,"overhead_gb":0.61,"total_gb":3.92,"total_gb_q8_cache":3.65},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":2.73,"kv_cache_gb":1.12,"overhead_gb":0.61,"total_gb":4.46,"total_gb_q8_cache":3.92},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":2.73,"kv_cache_gb":2.2,"overhead_gb":0.61,"total_gb":5.53,"total_gb_q8_cache":4.46},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":2.73,"kv_cache_gb":4.34,"overhead_gb":0.61,"total_gb":7.68,"total_gb_q8_cache":5.53},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":262144,"weights_gb":2.73,"kv_cache_gb":8.64,"overhead_gb":0.61,"total_gb":11.97,"total_gb_q8_cache":7.68},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":2.21,"kv_cache_gb":0.18,"overhead_gb":0.59,"total_gb":2.98,"total_gb_q8_cache":2.91},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":2.21,"kv_cache_gb":0.32,"overhead_gb":0.59,"total_gb":3.12,"total_gb_q8_cache":2.98},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":2.21,"kv_cache_gb":0.59,"overhead_gb":0.59,"total_gb":3.38,"total_gb_q8_cache":3.12},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":2.21,"kv_cache_gb":1.12,"overhead_gb":0.59,"total_gb":3.92,"total_gb_q8_cache":3.38},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":2.21,"kv_cache_gb":2.2,"overhead_gb":0.59,"total_gb":4.99,"total_gb_q8_cache":3.92},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":2.21,"kv_cache_gb":4.34,"overhead_gb":0.59,"total_gb":7.14,"total_gb_q8_cache":4.99},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":262144,"weights_gb":2.21,"kv_cache_gb":8.64,"overhead_gb":0.59,"total_gb":11.44,"total_gb_q8_cache":7.14},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":16,"kv_cache_gb":0.54,"overhead_gb":1.14,"total_gb":17.68,"total_gb_q8_cache":17.41},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":16,"kv_cache_gb":1.07,"overhead_gb":1.14,"total_gb":18.21,"total_gb_q8_cache":17.68},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":16,"kv_cache_gb":2.15,"overhead_gb":1.14,"total_gb":19.29,"total_gb_q8_cache":18.21},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":16,"kv_cache_gb":4.29,"overhead_gb":1.14,"total_gb":21.43,"total_gb_q8_cache":19.29},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":16,"kv_cache_gb":8.59,"overhead_gb":1.14,"total_gb":25.73,"total_gb_q8_cache":21.43},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":16,"kv_cache_gb":17.18,"overhead_gb":1.14,"total_gb":34.32,"total_gb_q8_cache":25.73},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":8.48,"kv_cache_gb":0.54,"overhead_gb":0.84,"total_gb":9.86,"total_gb_q8_cache":9.59},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":8.48,"kv_cache_gb":1.07,"overhead_gb":0.84,"total_gb":10.39,"total_gb_q8_cache":9.86},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":8.48,"kv_cache_gb":2.15,"overhead_gb":0.84,"total_gb":11.47,"total_gb_q8_cache":10.39},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":8.48,"kv_cache_gb":4.29,"overhead_gb":0.84,"total_gb":13.61,"total_gb_q8_cache":11.47},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":8.48,"kv_cache_gb":8.59,"overhead_gb":0.84,"total_gb":17.91,"total_gb_q8_cache":13.61},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":8.48,"kv_cache_gb":17.18,"overhead_gb":0.84,"total_gb":26.5,"total_gb_q8_cache":17.91},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":6.56,"kv_cache_gb":0.54,"overhead_gb":0.76,"total_gb":7.86,"total_gb_q8_cache":7.59},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":6.56,"kv_cache_gb":1.07,"overhead_gb":0.76,"total_gb":8.4,"total_gb_q8_cache":7.86},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":6.56,"kv_cache_gb":2.15,"overhead_gb":0.76,"total_gb":9.47,"total_gb_q8_cache":8.4},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":6.56,"kv_cache_gb":4.29,"overhead_gb":0.76,"total_gb":11.62,"total_gb_q8_cache":9.47},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":6.56,"kv_cache_gb":8.59,"overhead_gb":0.76,"total_gb":15.91,"total_gb_q8_cache":11.62},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":6.56,"kv_cache_gb":17.18,"overhead_gb":0.76,"total_gb":24.5,"total_gb_q8_cache":15.91},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":5.68,"kv_cache_gb":0.54,"overhead_gb":0.73,"total_gb":6.94,"total_gb_q8_cache":6.68},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":5.68,"kv_cache_gb":1.07,"overhead_gb":0.73,"total_gb":7.48,"total_gb_q8_cache":6.94},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":5.68,"kv_cache_gb":2.15,"overhead_gb":0.73,"total_gb":8.55,"total_gb_q8_cache":7.48},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":5.68,"kv_cache_gb":4.29,"overhead_gb":0.73,"total_gb":10.7,"total_gb_q8_cache":8.55},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":5.68,"kv_cache_gb":8.59,"overhead_gb":0.73,"total_gb":15,"total_gb_q8_cache":10.7},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":5.68,"kv_cache_gb":17.18,"overhead_gb":0.73,"total_gb":23.59,"total_gb_q8_cache":15},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":4.64,"kv_cache_gb":0.54,"overhead_gb":0.69,"total_gb":5.86,"total_gb_q8_cache":5.59},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":4.64,"kv_cache_gb":1.07,"overhead_gb":0.69,"total_gb":6.4,"total_gb_q8_cache":5.86},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":4.64,"kv_cache_gb":2.15,"overhead_gb":0.69,"total_gb":7.47,"total_gb_q8_cache":6.4},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":4.64,"kv_cache_gb":4.29,"overhead_gb":0.69,"total_gb":9.62,"total_gb_q8_cache":7.47},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":4.64,"kv_cache_gb":8.59,"overhead_gb":0.69,"total_gb":13.92,"total_gb_q8_cache":9.62},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":4.64,"kv_cache_gb":17.18,"overhead_gb":0.69,"total_gb":22.51,"total_gb_q8_cache":13.92},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":3.76,"kv_cache_gb":0.54,"overhead_gb":0.65,"total_gb":4.95,"total_gb_q8_cache":4.68},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":3.76,"kv_cache_gb":1.07,"overhead_gb":0.65,"total_gb":5.48,"total_gb_q8_cache":4.95},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":3.76,"kv_cache_gb":2.15,"overhead_gb":0.65,"total_gb":6.56,"total_gb_q8_cache":5.48},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":3.76,"kv_cache_gb":4.29,"overhead_gb":0.65,"total_gb":8.71,"total_gb_q8_cache":6.56},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":3.76,"kv_cache_gb":8.59,"overhead_gb":0.65,"total_gb":13,"total_gb_q8_cache":8.71},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":3.76,"kv_cache_gb":17.18,"overhead_gb":0.65,"total_gb":21.59,"total_gb_q8_cache":13},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":16.4,"kv_cache_gb":0.6,"overhead_gb":1.16,"total_gb":18.16,"total_gb_q8_cache":17.86},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":16.4,"kv_cache_gb":1.21,"overhead_gb":1.16,"total_gb":18.76,"total_gb_q8_cache":18.16},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":16.4,"kv_cache_gb":2.42,"overhead_gb":1.16,"total_gb":19.97,"total_gb_q8_cache":18.76},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":16.4,"kv_cache_gb":4.83,"overhead_gb":1.16,"total_gb":22.39,"total_gb_q8_cache":19.97},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"FP16","bytes_per_param":2,"context_tokens":40960,"weights_gb":16.4,"kv_cache_gb":6.04,"overhead_gb":1.16,"total_gb":23.6,"total_gb_q8_cache":20.58},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":8.69,"kv_cache_gb":0.6,"overhead_gb":0.85,"total_gb":10.14,"total_gb_q8_cache":9.84},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":8.69,"kv_cache_gb":1.21,"overhead_gb":0.85,"total_gb":10.75,"total_gb_q8_cache":10.14},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":8.69,"kv_cache_gb":2.42,"overhead_gb":0.85,"total_gb":11.96,"total_gb_q8_cache":10.75},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":8.69,"kv_cache_gb":4.83,"overhead_gb":0.85,"total_gb":14.37,"total_gb_q8_cache":11.96},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":40960,"weights_gb":8.69,"kv_cache_gb":6.04,"overhead_gb":0.85,"total_gb":15.58,"total_gb_q8_cache":12.56},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":6.72,"kv_cache_gb":0.6,"overhead_gb":0.77,"total_gb":8.1,"total_gb_q8_cache":7.79},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":6.72,"kv_cache_gb":1.21,"overhead_gb":0.77,"total_gb":8.7,"total_gb_q8_cache":8.1},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":6.72,"kv_cache_gb":2.42,"overhead_gb":0.77,"total_gb":9.91,"total_gb_q8_cache":8.7},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":6.72,"kv_cache_gb":4.83,"overhead_gb":0.77,"total_gb":12.32,"total_gb_q8_cache":9.91},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":40960,"weights_gb":6.72,"kv_cache_gb":6.04,"overhead_gb":0.77,"total_gb":13.53,"total_gb_q8_cache":10.51},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":5.82,"kv_cache_gb":0.6,"overhead_gb":0.73,"total_gb":7.16,"total_gb_q8_cache":6.86},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":5.82,"kv_cache_gb":1.21,"overhead_gb":0.73,"total_gb":7.76,"total_gb_q8_cache":7.16},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":5.82,"kv_cache_gb":2.42,"overhead_gb":0.73,"total_gb":8.97,"total_gb_q8_cache":7.76},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":5.82,"kv_cache_gb":4.83,"overhead_gb":0.73,"total_gb":11.39,"total_gb_q8_cache":8.97},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":40960,"weights_gb":5.82,"kv_cache_gb":6.04,"overhead_gb":0.73,"total_gb":12.59,"total_gb_q8_cache":9.57},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":4.76,"kv_cache_gb":0.6,"overhead_gb":0.69,"total_gb":6.05,"total_gb_q8_cache":5.75},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":4.76,"kv_cache_gb":1.21,"overhead_gb":0.69,"total_gb":6.65,"total_gb_q8_cache":6.05},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":4.76,"kv_cache_gb":2.42,"overhead_gb":0.69,"total_gb":7.86,"total_gb_q8_cache":6.65},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":4.76,"kv_cache_gb":4.83,"overhead_gb":0.69,"total_gb":10.28,"total_gb_q8_cache":7.86},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":40960,"weights_gb":4.76,"kv_cache_gb":6.04,"overhead_gb":0.69,"total_gb":11.49,"total_gb_q8_cache":8.47},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":3.85,"kv_cache_gb":0.6,"overhead_gb":0.65,"total_gb":5.11,"total_gb_q8_cache":4.81},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":3.85,"kv_cache_gb":1.21,"overhead_gb":0.65,"total_gb":5.72,"total_gb_q8_cache":5.11},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":3.85,"kv_cache_gb":2.42,"overhead_gb":0.65,"total_gb":6.92,"total_gb_q8_cache":5.72},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":3.85,"kv_cache_gb":4.83,"overhead_gb":0.65,"total_gb":9.34,"total_gb_q8_cache":6.92},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":40960,"weights_gb":3.85,"kv_cache_gb":6.04,"overhead_gb":0.65,"total_gb":10.55,"total_gb_q8_cache":7.53},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":19.4,"kv_cache_gb":0.18,"overhead_gb":1.28,"total_gb":20.86,"total_gb_q8_cache":20.79},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":19.4,"kv_cache_gb":0.32,"overhead_gb":1.28,"total_gb":20.99,"total_gb_q8_cache":20.86},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":19.4,"kv_cache_gb":0.59,"overhead_gb":1.28,"total_gb":21.26,"total_gb_q8_cache":20.99},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":19.4,"kv_cache_gb":1.12,"overhead_gb":1.28,"total_gb":21.8,"total_gb_q8_cache":21.26},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":19.4,"kv_cache_gb":2.2,"overhead_gb":1.28,"total_gb":22.87,"total_gb_q8_cache":21.8},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":19.4,"kv_cache_gb":4.34,"overhead_gb":1.28,"total_gb":25.02,"total_gb_q8_cache":22.87},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"FP16","bytes_per_param":2,"context_tokens":262144,"weights_gb":19.4,"kv_cache_gb":8.64,"overhead_gb":1.28,"total_gb":29.32,"total_gb_q8_cache":25.02},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":10.28,"kv_cache_gb":0.18,"overhead_gb":0.91,"total_gb":11.38,"total_gb_q8_cache":11.31},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":10.28,"kv_cache_gb":0.32,"overhead_gb":0.91,"total_gb":11.51,"total_gb_q8_cache":11.38},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":10.28,"kv_cache_gb":0.59,"overhead_gb":0.91,"total_gb":11.78,"total_gb_q8_cache":11.51},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":10.28,"kv_cache_gb":1.12,"overhead_gb":0.91,"total_gb":12.32,"total_gb_q8_cache":11.78},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":10.28,"kv_cache_gb":2.2,"overhead_gb":0.91,"total_gb":13.39,"total_gb_q8_cache":12.32},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":10.28,"kv_cache_gb":4.34,"overhead_gb":0.91,"total_gb":15.54,"total_gb_q8_cache":13.39},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":262144,"weights_gb":10.28,"kv_cache_gb":8.64,"overhead_gb":0.91,"total_gb":19.83,"total_gb_q8_cache":15.54},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":7.95,"kv_cache_gb":0.18,"overhead_gb":0.82,"total_gb":8.96,"total_gb_q8_cache":8.89},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":7.95,"kv_cache_gb":0.32,"overhead_gb":0.82,"total_gb":9.09,"total_gb_q8_cache":8.96},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":7.95,"kv_cache_gb":0.59,"overhead_gb":0.82,"total_gb":9.36,"total_gb_q8_cache":9.09},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":7.95,"kv_cache_gb":1.12,"overhead_gb":0.82,"total_gb":9.9,"total_gb_q8_cache":9.36},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":7.95,"kv_cache_gb":2.2,"overhead_gb":0.82,"total_gb":10.97,"total_gb_q8_cache":9.9},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":7.95,"kv_cache_gb":4.34,"overhead_gb":0.82,"total_gb":13.12,"total_gb_q8_cache":10.97},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":262144,"weights_gb":7.95,"kv_cache_gb":8.64,"overhead_gb":0.82,"total_gb":17.41,"total_gb_q8_cache":13.12},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":6.89,"kv_cache_gb":0.18,"overhead_gb":0.78,"total_gb":7.85,"total_gb_q8_cache":7.78},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":6.89,"kv_cache_gb":0.32,"overhead_gb":0.78,"total_gb":7.98,"total_gb_q8_cache":7.85},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":6.89,"kv_cache_gb":0.59,"overhead_gb":0.78,"total_gb":8.25,"total_gb_q8_cache":7.98},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":6.89,"kv_cache_gb":1.12,"overhead_gb":0.78,"total_gb":8.79,"total_gb_q8_cache":8.25},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":6.89,"kv_cache_gb":2.2,"overhead_gb":0.78,"total_gb":9.86,"total_gb_q8_cache":8.79},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":6.89,"kv_cache_gb":4.34,"overhead_gb":0.78,"total_gb":12.01,"total_gb_q8_cache":9.86},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":262144,"weights_gb":6.89,"kv_cache_gb":8.64,"overhead_gb":0.78,"total_gb":16.3,"total_gb_q8_cache":12.01},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":5.63,"kv_cache_gb":0.18,"overhead_gb":0.73,"total_gb":6.54,"total_gb_q8_cache":6.47},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":5.63,"kv_cache_gb":0.32,"overhead_gb":0.73,"total_gb":6.67,"total_gb_q8_cache":6.54},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":5.63,"kv_cache_gb":0.59,"overhead_gb":0.73,"total_gb":6.94,"total_gb_q8_cache":6.67},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":5.63,"kv_cache_gb":1.12,"overhead_gb":0.73,"total_gb":7.47,"total_gb_q8_cache":6.94},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":5.63,"kv_cache_gb":2.2,"overhead_gb":0.73,"total_gb":8.55,"total_gb_q8_cache":7.47},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":5.63,"kv_cache_gb":4.34,"overhead_gb":0.73,"total_gb":10.7,"total_gb_q8_cache":8.55},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":262144,"weights_gb":5.63,"kv_cache_gb":8.64,"overhead_gb":0.73,"total_gb":14.99,"total_gb_q8_cache":10.7},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":4.56,"kv_cache_gb":0.18,"overhead_gb":0.68,"total_gb":5.43,"total_gb_q8_cache":5.36},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":4.56,"kv_cache_gb":0.32,"overhead_gb":0.68,"total_gb":5.56,"total_gb_q8_cache":5.43},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":4.56,"kv_cache_gb":0.59,"overhead_gb":0.68,"total_gb":5.83,"total_gb_q8_cache":5.56},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":4.56,"kv_cache_gb":1.12,"overhead_gb":0.68,"total_gb":6.37,"total_gb_q8_cache":5.83},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":4.56,"kv_cache_gb":2.2,"overhead_gb":0.68,"total_gb":7.44,"total_gb_q8_cache":6.37},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":4.56,"kv_cache_gb":4.34,"overhead_gb":0.68,"total_gb":9.59,"total_gb_q8_cache":7.44},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":262144,"weights_gb":4.56,"kv_cache_gb":8.64,"overhead_gb":0.68,"total_gb":13.88,"total_gb_q8_cache":9.59},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":24,"kv_cache_gb":0.4,"overhead_gb":1.46,"total_gb":25.86,"total_gb_q8_cache":25.66},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":24,"kv_cache_gb":0.47,"overhead_gb":1.46,"total_gb":25.93,"total_gb_q8_cache":25.69},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":24,"kv_cache_gb":0.6,"overhead_gb":1.46,"total_gb":26.06,"total_gb_q8_cache":25.76},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":24,"kv_cache_gb":0.87,"overhead_gb":1.46,"total_gb":26.33,"total_gb_q8_cache":25.9},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":24,"kv_cache_gb":1.41,"overhead_gb":1.46,"total_gb":26.87,"total_gb_q8_cache":26.16},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":24,"kv_cache_gb":2.48,"overhead_gb":1.46,"total_gb":27.94,"total_gb_q8_cache":26.7},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"FP16","bytes_per_param":2,"context_tokens":262144,"weights_gb":24,"kv_cache_gb":4.63,"overhead_gb":1.46,"total_gb":30.09,"total_gb_q8_cache":27.78},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":12.72,"kv_cache_gb":0.4,"overhead_gb":1.01,"total_gb":14.13,"total_gb_q8_cache":13.93},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":12.72,"kv_cache_gb":0.47,"overhead_gb":1.01,"total_gb":14.2,"total_gb_q8_cache":13.96},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":12.72,"kv_cache_gb":0.6,"overhead_gb":1.01,"total_gb":14.33,"total_gb_q8_cache":14.03},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":12.72,"kv_cache_gb":0.87,"overhead_gb":1.01,"total_gb":14.6,"total_gb_q8_cache":14.17},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":12.72,"kv_cache_gb":1.41,"overhead_gb":1.01,"total_gb":15.14,"total_gb_q8_cache":14.43},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":12.72,"kv_cache_gb":2.48,"overhead_gb":1.01,"total_gb":16.21,"total_gb_q8_cache":14.97},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":262144,"weights_gb":12.72,"kv_cache_gb":4.63,"overhead_gb":1.01,"total_gb":18.36,"total_gb_q8_cache":16.04},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":9.84,"kv_cache_gb":0.4,"overhead_gb":0.89,"total_gb":11.14,"total_gb_q8_cache":10.93},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":9.84,"kv_cache_gb":0.47,"overhead_gb":0.89,"total_gb":11.2,"total_gb_q8_cache":10.97},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":9.84,"kv_cache_gb":0.6,"overhead_gb":0.89,"total_gb":11.34,"total_gb_q8_cache":11.04},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":9.84,"kv_cache_gb":0.87,"overhead_gb":0.89,"total_gb":11.61,"total_gb_q8_cache":11.17},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":9.84,"kv_cache_gb":1.41,"overhead_gb":0.89,"total_gb":12.14,"total_gb_q8_cache":11.44},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":9.84,"kv_cache_gb":2.48,"overhead_gb":0.89,"total_gb":13.22,"total_gb_q8_cache":11.98},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":262144,"weights_gb":9.84,"kv_cache_gb":4.63,"overhead_gb":0.89,"total_gb":15.36,"total_gb_q8_cache":13.05},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":8.52,"kv_cache_gb":0.4,"overhead_gb":0.84,"total_gb":9.76,"total_gb_q8_cache":9.56},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":8.52,"kv_cache_gb":0.47,"overhead_gb":0.84,"total_gb":9.83,"total_gb_q8_cache":9.6},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":8.52,"kv_cache_gb":0.6,"overhead_gb":0.84,"total_gb":9.96,"total_gb_q8_cache":9.66},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":8.52,"kv_cache_gb":0.87,"overhead_gb":0.84,"total_gb":10.23,"total_gb_q8_cache":9.8},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":8.52,"kv_cache_gb":1.41,"overhead_gb":0.84,"total_gb":10.77,"total_gb_q8_cache":10.07},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":8.52,"kv_cache_gb":2.48,"overhead_gb":0.84,"total_gb":11.84,"total_gb_q8_cache":10.6},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":262144,"weights_gb":8.52,"kv_cache_gb":4.63,"overhead_gb":0.84,"total_gb":13.99,"total_gb_q8_cache":11.68},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":6.96,"kv_cache_gb":0.4,"overhead_gb":0.78,"total_gb":8.14,"total_gb_q8_cache":7.94},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":6.96,"kv_cache_gb":0.47,"overhead_gb":0.78,"total_gb":8.21,"total_gb_q8_cache":7.97},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":6.96,"kv_cache_gb":0.6,"overhead_gb":0.78,"total_gb":8.34,"total_gb_q8_cache":8.04},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":6.96,"kv_cache_gb":0.87,"overhead_gb":0.78,"total_gb":8.61,"total_gb_q8_cache":8.17},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":6.96,"kv_cache_gb":1.41,"overhead_gb":0.78,"total_gb":9.15,"total_gb_q8_cache":8.44},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":6.96,"kv_cache_gb":2.48,"overhead_gb":0.78,"total_gb":10.22,"total_gb_q8_cache":8.98},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":262144,"weights_gb":6.96,"kv_cache_gb":4.63,"overhead_gb":0.78,"total_gb":12.37,"total_gb_q8_cache":10.05},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":5.64,"kv_cache_gb":0.4,"overhead_gb":0.73,"total_gb":6.77,"total_gb_q8_cache":6.57},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":5.64,"kv_cache_gb":0.47,"overhead_gb":0.73,"total_gb":6.84,"total_gb_q8_cache":6.6},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":5.64,"kv_cache_gb":0.6,"overhead_gb":0.73,"total_gb":6.97,"total_gb_q8_cache":6.67},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":5.64,"kv_cache_gb":0.87,"overhead_gb":0.73,"total_gb":7.24,"total_gb_q8_cache":6.8},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":5.64,"kv_cache_gb":1.41,"overhead_gb":0.73,"total_gb":7.77,"total_gb_q8_cache":7.07},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":5.64,"kv_cache_gb":2.48,"overhead_gb":0.73,"total_gb":8.85,"total_gb_q8_cache":7.61},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":262144,"weights_gb":5.64,"kv_cache_gb":4.63,"overhead_gb":0.73,"total_gb":11,"total_gb_q8_cache":8.68},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":24.4,"kv_cache_gb":0.6,"overhead_gb":1.48,"total_gb":26.48,"total_gb_q8_cache":26.18},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":24.4,"kv_cache_gb":0.87,"overhead_gb":1.48,"total_gb":26.75,"total_gb_q8_cache":26.31},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":24.4,"kv_cache_gb":1.41,"overhead_gb":1.48,"total_gb":27.29,"total_gb_q8_cache":26.58},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":24.4,"kv_cache_gb":2.48,"overhead_gb":1.48,"total_gb":28.36,"total_gb_q8_cache":27.12},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":24.4,"kv_cache_gb":4.63,"overhead_gb":1.48,"total_gb":30.51,"total_gb_q8_cache":28.19},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":24.4,"kv_cache_gb":8.93,"overhead_gb":1.48,"total_gb":34.8,"total_gb_q8_cache":30.34},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":12.93,"kv_cache_gb":0.6,"overhead_gb":1.02,"total_gb":14.55,"total_gb_q8_cache":14.25},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":12.93,"kv_cache_gb":0.87,"overhead_gb":1.02,"total_gb":14.82,"total_gb_q8_cache":14.39},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":12.93,"kv_cache_gb":1.41,"overhead_gb":1.02,"total_gb":15.36,"total_gb_q8_cache":14.65},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":12.93,"kv_cache_gb":2.48,"overhead_gb":1.02,"total_gb":16.43,"total_gb_q8_cache":15.19},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":12.93,"kv_cache_gb":4.63,"overhead_gb":1.02,"total_gb":18.58,"total_gb_q8_cache":16.26},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":12.93,"kv_cache_gb":8.93,"overhead_gb":1.02,"total_gb":22.87,"total_gb_q8_cache":18.41},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":10,"kv_cache_gb":0.6,"overhead_gb":0.9,"total_gb":11.51,"total_gb_q8_cache":11.21},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":10,"kv_cache_gb":0.87,"overhead_gb":0.9,"total_gb":11.78,"total_gb_q8_cache":11.34},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":10,"kv_cache_gb":1.41,"overhead_gb":0.9,"total_gb":12.31,"total_gb_q8_cache":11.61},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":10,"kv_cache_gb":2.48,"overhead_gb":0.9,"total_gb":13.39,"total_gb_q8_cache":12.15},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":10,"kv_cache_gb":4.63,"overhead_gb":0.9,"total_gb":15.53,"total_gb_q8_cache":13.22},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":10,"kv_cache_gb":8.93,"overhead_gb":0.9,"total_gb":19.83,"total_gb_q8_cache":15.37},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":8.66,"kv_cache_gb":0.6,"overhead_gb":0.85,"total_gb":10.11,"total_gb_q8_cache":9.81},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":8.66,"kv_cache_gb":0.87,"overhead_gb":0.85,"total_gb":10.38,"total_gb_q8_cache":9.94},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":8.66,"kv_cache_gb":1.41,"overhead_gb":0.85,"total_gb":10.92,"total_gb_q8_cache":10.21},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":8.66,"kv_cache_gb":2.48,"overhead_gb":0.85,"total_gb":11.99,"total_gb_q8_cache":10.75},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":8.66,"kv_cache_gb":4.63,"overhead_gb":0.85,"total_gb":14.14,"total_gb_q8_cache":11.82},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":8.66,"kv_cache_gb":8.93,"overhead_gb":0.85,"total_gb":18.43,"total_gb_q8_cache":13.97},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":7.08,"kv_cache_gb":0.6,"overhead_gb":0.78,"total_gb":8.46,"total_gb_q8_cache":8.16},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":7.08,"kv_cache_gb":0.87,"overhead_gb":0.78,"total_gb":8.73,"total_gb_q8_cache":8.3},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":7.08,"kv_cache_gb":1.41,"overhead_gb":0.78,"total_gb":9.27,"total_gb_q8_cache":8.56},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":7.08,"kv_cache_gb":2.48,"overhead_gb":0.78,"total_gb":10.34,"total_gb_q8_cache":9.1},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":7.08,"kv_cache_gb":4.63,"overhead_gb":0.78,"total_gb":12.49,"total_gb_q8_cache":10.17},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":7.08,"kv_cache_gb":8.93,"overhead_gb":0.78,"total_gb":16.78,"total_gb_q8_cache":12.32},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":5.73,"kv_cache_gb":0.6,"overhead_gb":0.73,"total_gb":7.07,"total_gb_q8_cache":6.77},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":5.73,"kv_cache_gb":0.87,"overhead_gb":0.73,"total_gb":7.34,"total_gb_q8_cache":6.9},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":5.73,"kv_cache_gb":1.41,"overhead_gb":0.73,"total_gb":7.87,"total_gb_q8_cache":7.17},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":5.73,"kv_cache_gb":2.48,"overhead_gb":0.73,"total_gb":8.95,"total_gb_q8_cache":7.7},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":5.73,"kv_cache_gb":4.63,"overhead_gb":0.73,"total_gb":11.09,"total_gb_q8_cache":8.78},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":5.73,"kv_cache_gb":8.93,"overhead_gb":0.73,"total_gb":15.39,"total_gb_q8_cache":10.93},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":28,"kv_cache_gb":0.67,"overhead_gb":1.62,"total_gb":30.29,"total_gb_q8_cache":29.96},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":28,"kv_cache_gb":1.34,"overhead_gb":1.62,"total_gb":30.96,"total_gb_q8_cache":30.29},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":28,"kv_cache_gb":2.68,"overhead_gb":1.62,"total_gb":32.3,"total_gb_q8_cache":30.96},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":28,"kv_cache_gb":5.37,"overhead_gb":1.62,"total_gb":34.99,"total_gb_q8_cache":32.3},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":28,"kv_cache_gb":10.74,"overhead_gb":1.62,"total_gb":40.36,"total_gb_q8_cache":34.99},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":28,"kv_cache_gb":21.47,"overhead_gb":1.62,"total_gb":51.09,"total_gb_q8_cache":40.36},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"FP16","bytes_per_param":2,"context_tokens":262144,"weights_gb":28,"kv_cache_gb":42.95,"overhead_gb":1.62,"total_gb":72.57,"total_gb_q8_cache":51.09},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":14.84,"kv_cache_gb":0.67,"overhead_gb":1.09,"total_gb":16.6,"total_gb_q8_cache":16.27},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":14.84,"kv_cache_gb":1.34,"overhead_gb":1.09,"total_gb":17.28,"total_gb_q8_cache":16.6},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":14.84,"kv_cache_gb":2.68,"overhead_gb":1.09,"total_gb":18.62,"total_gb_q8_cache":17.28},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":14.84,"kv_cache_gb":5.37,"overhead_gb":1.09,"total_gb":21.3,"total_gb_q8_cache":18.62},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":14.84,"kv_cache_gb":10.74,"overhead_gb":1.09,"total_gb":26.67,"total_gb_q8_cache":21.3},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":14.84,"kv_cache_gb":21.47,"overhead_gb":1.09,"total_gb":37.41,"total_gb_q8_cache":26.67},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":262144,"weights_gb":14.84,"kv_cache_gb":42.95,"overhead_gb":1.09,"total_gb":58.88,"total_gb_q8_cache":37.41},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":11.48,"kv_cache_gb":0.67,"overhead_gb":0.96,"total_gb":13.11,"total_gb_q8_cache":12.77},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":11.48,"kv_cache_gb":1.34,"overhead_gb":0.96,"total_gb":13.78,"total_gb_q8_cache":13.11},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":11.48,"kv_cache_gb":2.68,"overhead_gb":0.96,"total_gb":15.12,"total_gb_q8_cache":13.78},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":11.48,"kv_cache_gb":5.37,"overhead_gb":0.96,"total_gb":17.81,"total_gb_q8_cache":15.12},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":11.48,"kv_cache_gb":10.74,"overhead_gb":0.96,"total_gb":23.18,"total_gb_q8_cache":17.81},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":11.48,"kv_cache_gb":21.47,"overhead_gb":0.96,"total_gb":33.91,"total_gb_q8_cache":23.18},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":262144,"weights_gb":11.48,"kv_cache_gb":42.95,"overhead_gb":0.96,"total_gb":55.39,"total_gb_q8_cache":33.91},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":9.94,"kv_cache_gb":0.67,"overhead_gb":0.9,"total_gb":11.51,"total_gb_q8_cache":11.17},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":9.94,"kv_cache_gb":1.34,"overhead_gb":0.9,"total_gb":12.18,"total_gb_q8_cache":11.51},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":9.94,"kv_cache_gb":2.68,"overhead_gb":0.9,"total_gb":13.52,"total_gb_q8_cache":12.18},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":9.94,"kv_cache_gb":5.37,"overhead_gb":0.9,"total_gb":16.21,"total_gb_q8_cache":13.52},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":9.94,"kv_cache_gb":10.74,"overhead_gb":0.9,"total_gb":21.58,"total_gb_q8_cache":16.21},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":9.94,"kv_cache_gb":21.47,"overhead_gb":0.9,"total_gb":32.31,"total_gb_q8_cache":21.58},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":262144,"weights_gb":9.94,"kv_cache_gb":42.95,"overhead_gb":0.9,"total_gb":53.79,"total_gb_q8_cache":32.31},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":8.12,"kv_cache_gb":0.67,"overhead_gb":0.82,"total_gb":9.62,"total_gb_q8_cache":9.28},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":8.12,"kv_cache_gb":1.34,"overhead_gb":0.82,"total_gb":10.29,"total_gb_q8_cache":9.62},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":8.12,"kv_cache_gb":2.68,"overhead_gb":0.82,"total_gb":11.63,"total_gb_q8_cache":10.29},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":8.12,"kv_cache_gb":5.37,"overhead_gb":0.82,"total_gb":14.31,"total_gb_q8_cache":11.63},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":8.12,"kv_cache_gb":10.74,"overhead_gb":0.82,"total_gb":19.68,"total_gb_q8_cache":14.31},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":8.12,"kv_cache_gb":21.47,"overhead_gb":0.82,"total_gb":30.42,"total_gb_q8_cache":19.68},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":262144,"weights_gb":8.12,"kv_cache_gb":42.95,"overhead_gb":0.82,"total_gb":51.89,"total_gb_q8_cache":30.42},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":6.58,"kv_cache_gb":0.67,"overhead_gb":0.76,"total_gb":8.01,"total_gb_q8_cache":7.68},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":6.58,"kv_cache_gb":1.34,"overhead_gb":0.76,"total_gb":8.69,"total_gb_q8_cache":8.01},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":6.58,"kv_cache_gb":2.68,"overhead_gb":0.76,"total_gb":10.03,"total_gb_q8_cache":8.69},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":6.58,"kv_cache_gb":5.37,"overhead_gb":0.76,"total_gb":12.71,"total_gb_q8_cache":10.03},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":6.58,"kv_cache_gb":10.74,"overhead_gb":0.76,"total_gb":18.08,"total_gb_q8_cache":12.71},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":6.58,"kv_cache_gb":21.47,"overhead_gb":0.76,"total_gb":28.82,"total_gb_q8_cache":18.08},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":262144,"weights_gb":6.58,"kv_cache_gb":42.95,"overhead_gb":0.76,"total_gb":50.29,"total_gb_q8_cache":28.82},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":29.6,"kv_cache_gb":0.67,"overhead_gb":1.68,"total_gb":31.96,"total_gb_q8_cache":31.62},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":29.6,"kv_cache_gb":1.34,"overhead_gb":1.68,"total_gb":32.63,"total_gb_q8_cache":31.96},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":29.6,"kv_cache_gb":2.68,"overhead_gb":1.68,"total_gb":33.97,"total_gb_q8_cache":32.63},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":29.6,"kv_cache_gb":5.37,"overhead_gb":1.68,"total_gb":36.65,"total_gb_q8_cache":33.97},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"FP16","bytes_per_param":2,"context_tokens":40960,"weights_gb":29.6,"kv_cache_gb":6.71,"overhead_gb":1.68,"total_gb":37.99,"total_gb_q8_cache":34.64},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":15.69,"kv_cache_gb":0.67,"overhead_gb":1.13,"total_gb":17.49,"total_gb_q8_cache":17.15},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":15.69,"kv_cache_gb":1.34,"overhead_gb":1.13,"total_gb":18.16,"total_gb_q8_cache":17.49},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":15.69,"kv_cache_gb":2.68,"overhead_gb":1.13,"total_gb":19.5,"total_gb_q8_cache":18.16},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":15.69,"kv_cache_gb":5.37,"overhead_gb":1.13,"total_gb":22.18,"total_gb_q8_cache":19.5},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":40960,"weights_gb":15.69,"kv_cache_gb":6.71,"overhead_gb":1.13,"total_gb":23.53,"total_gb_q8_cache":20.17},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":12.14,"kv_cache_gb":0.67,"overhead_gb":0.99,"total_gb":13.79,"total_gb_q8_cache":13.46},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":12.14,"kv_cache_gb":1.34,"overhead_gb":0.99,"total_gb":14.46,"total_gb_q8_cache":13.79},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":12.14,"kv_cache_gb":2.68,"overhead_gb":0.99,"total_gb":15.81,"total_gb_q8_cache":14.46},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":12.14,"kv_cache_gb":5.37,"overhead_gb":0.99,"total_gb":18.49,"total_gb_q8_cache":15.81},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":40960,"weights_gb":12.14,"kv_cache_gb":6.71,"overhead_gb":0.99,"total_gb":19.83,"total_gb_q8_cache":16.48},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":10.51,"kv_cache_gb":0.67,"overhead_gb":0.92,"total_gb":12.1,"total_gb_q8_cache":11.76},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":10.51,"kv_cache_gb":1.34,"overhead_gb":0.92,"total_gb":12.77,"total_gb_q8_cache":12.1},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":10.51,"kv_cache_gb":2.68,"overhead_gb":0.92,"total_gb":14.11,"total_gb_q8_cache":12.77},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":10.51,"kv_cache_gb":5.37,"overhead_gb":0.92,"total_gb":16.8,"total_gb_q8_cache":14.11},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":40960,"weights_gb":10.51,"kv_cache_gb":6.71,"overhead_gb":0.92,"total_gb":18.14,"total_gb_q8_cache":14.78},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":8.58,"kv_cache_gb":0.67,"overhead_gb":0.84,"total_gb":10.1,"total_gb_q8_cache":9.76},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":8.58,"kv_cache_gb":1.34,"overhead_gb":0.84,"total_gb":10.77,"total_gb_q8_cache":10.1},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":8.58,"kv_cache_gb":2.68,"overhead_gb":0.84,"total_gb":12.11,"total_gb_q8_cache":10.77},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":8.58,"kv_cache_gb":5.37,"overhead_gb":0.84,"total_gb":14.8,"total_gb_q8_cache":12.11},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":40960,"weights_gb":8.58,"kv_cache_gb":6.71,"overhead_gb":0.84,"total_gb":16.14,"total_gb_q8_cache":12.78},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":6.96,"kv_cache_gb":0.67,"overhead_gb":0.78,"total_gb":8.41,"total_gb_q8_cache":8.07},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":6.96,"kv_cache_gb":1.34,"overhead_gb":0.78,"total_gb":9.08,"total_gb_q8_cache":8.41},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":6.96,"kv_cache_gb":2.68,"overhead_gb":0.78,"total_gb":10.42,"total_gb_q8_cache":9.08},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":6.96,"kv_cache_gb":5.37,"overhead_gb":0.78,"total_gb":13.1,"total_gb_q8_cache":10.42},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":40960,"weights_gb":6.96,"kv_cache_gb":6.71,"overhead_gb":0.78,"total_gb":14.45,"total_gb_q8_cache":11.09},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":29.4,"kv_cache_gb":0.84,"overhead_gb":1.68,"total_gb":31.91,"total_gb_q8_cache":31.5},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":29.4,"kv_cache_gb":1.68,"overhead_gb":1.68,"total_gb":32.75,"total_gb_q8_cache":31.91},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":29.4,"kv_cache_gb":3.36,"overhead_gb":1.68,"total_gb":34.43,"total_gb_q8_cache":32.75},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":15.58,"kv_cache_gb":0.84,"overhead_gb":1.12,"total_gb":17.54,"total_gb_q8_cache":17.12},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":15.58,"kv_cache_gb":1.68,"overhead_gb":1.12,"total_gb":18.38,"total_gb_q8_cache":17.54},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":15.58,"kv_cache_gb":3.36,"overhead_gb":1.12,"total_gb":20.06,"total_gb_q8_cache":18.38},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":12.05,"kv_cache_gb":0.84,"overhead_gb":0.98,"total_gb":13.88,"total_gb_q8_cache":13.46},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":12.05,"kv_cache_gb":1.68,"overhead_gb":0.98,"total_gb":14.71,"total_gb_q8_cache":13.88},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":12.05,"kv_cache_gb":3.36,"overhead_gb":0.98,"total_gb":16.39,"total_gb_q8_cache":14.71},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":10.44,"kv_cache_gb":0.84,"overhead_gb":0.92,"total_gb":12.19,"total_gb_q8_cache":11.77},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":10.44,"kv_cache_gb":1.68,"overhead_gb":0.92,"total_gb":13.03,"total_gb_q8_cache":12.19},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":10.44,"kv_cache_gb":3.36,"overhead_gb":0.92,"total_gb":14.71,"total_gb_q8_cache":13.03},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":8.53,"kv_cache_gb":0.84,"overhead_gb":0.84,"total_gb":10.21,"total_gb_q8_cache":9.79},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":8.53,"kv_cache_gb":1.68,"overhead_gb":0.84,"total_gb":11.04,"total_gb_q8_cache":10.21},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":8.53,"kv_cache_gb":3.36,"overhead_gb":0.84,"total_gb":12.72,"total_gb_q8_cache":11.04},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":6.91,"kv_cache_gb":0.84,"overhead_gb":0.78,"total_gb":8.52,"total_gb_q8_cache":8.1},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":6.91,"kv_cache_gb":1.68,"overhead_gb":0.78,"total_gb":9.36,"total_gb_q8_cache":8.52},
    {"model_id":"phi-4","model":"Phi-4 14B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":6.91,"kv_cache_gb":3.36,"overhead_gb":0.78,"total_gb":11.04,"total_gb_q8_cache":9.36},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":42,"kv_cache_gb":0.1,"overhead_gb":2.18,"total_gb":44.28,"total_gb_q8_cache":44.23},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":42,"kv_cache_gb":0.2,"overhead_gb":2.18,"total_gb":44.38,"total_gb_q8_cache":44.28},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":42,"kv_cache_gb":0.41,"overhead_gb":2.18,"total_gb":44.59,"total_gb_q8_cache":44.38},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":42,"kv_cache_gb":0.81,"overhead_gb":2.18,"total_gb":44.99,"total_gb_q8_cache":44.58},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":42,"kv_cache_gb":1.61,"overhead_gb":2.18,"total_gb":45.79,"total_gb_q8_cache":44.99},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":42,"kv_cache_gb":3.22,"overhead_gb":2.18,"total_gb":47.4,"total_gb_q8_cache":45.79},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":22.26,"kv_cache_gb":0.1,"overhead_gb":1.39,"total_gb":23.75,"total_gb_q8_cache":23.7},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":22.26,"kv_cache_gb":0.2,"overhead_gb":1.39,"total_gb":23.85,"total_gb_q8_cache":23.75},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":22.26,"kv_cache_gb":0.41,"overhead_gb":1.39,"total_gb":24.06,"total_gb_q8_cache":23.85},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":22.26,"kv_cache_gb":0.81,"overhead_gb":1.39,"total_gb":24.46,"total_gb_q8_cache":24.05},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":22.26,"kv_cache_gb":1.61,"overhead_gb":1.39,"total_gb":25.26,"total_gb_q8_cache":24.46},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":22.26,"kv_cache_gb":3.22,"overhead_gb":1.39,"total_gb":26.87,"total_gb_q8_cache":25.26},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":17.22,"kv_cache_gb":0.1,"overhead_gb":1.19,"total_gb":18.51,"total_gb_q8_cache":18.46},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":17.22,"kv_cache_gb":0.2,"overhead_gb":1.19,"total_gb":18.61,"total_gb_q8_cache":18.51},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":17.22,"kv_cache_gb":0.41,"overhead_gb":1.19,"total_gb":18.81,"total_gb_q8_cache":18.61},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":17.22,"kv_cache_gb":0.81,"overhead_gb":1.19,"total_gb":19.22,"total_gb_q8_cache":18.81},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":17.22,"kv_cache_gb":1.61,"overhead_gb":1.19,"total_gb":20.02,"total_gb_q8_cache":19.22},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":17.22,"kv_cache_gb":3.22,"overhead_gb":1.19,"total_gb":21.63,"total_gb_q8_cache":20.02},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":14.91,"kv_cache_gb":0.1,"overhead_gb":1.1,"total_gb":16.11,"total_gb_q8_cache":16.06},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":14.91,"kv_cache_gb":0.2,"overhead_gb":1.1,"total_gb":16.21,"total_gb_q8_cache":16.11},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":14.91,"kv_cache_gb":0.41,"overhead_gb":1.1,"total_gb":16.41,"total_gb_q8_cache":16.21},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":14.91,"kv_cache_gb":0.81,"overhead_gb":1.1,"total_gb":16.81,"total_gb_q8_cache":16.41},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":14.91,"kv_cache_gb":1.61,"overhead_gb":1.1,"total_gb":17.62,"total_gb_q8_cache":16.81},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":14.91,"kv_cache_gb":3.22,"overhead_gb":1.1,"total_gb":19.23,"total_gb_q8_cache":17.62},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":12.18,"kv_cache_gb":0.1,"overhead_gb":0.99,"total_gb":13.27,"total_gb_q8_cache":13.22},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":12.18,"kv_cache_gb":0.2,"overhead_gb":0.99,"total_gb":13.37,"total_gb_q8_cache":13.27},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":12.18,"kv_cache_gb":0.41,"overhead_gb":0.99,"total_gb":13.57,"total_gb_q8_cache":13.37},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":12.18,"kv_cache_gb":0.81,"overhead_gb":0.99,"total_gb":13.98,"total_gb_q8_cache":13.57},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":12.18,"kv_cache_gb":1.61,"overhead_gb":0.99,"total_gb":14.78,"total_gb_q8_cache":13.97},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":12.18,"kv_cache_gb":3.22,"overhead_gb":0.99,"total_gb":16.39,"total_gb_q8_cache":14.78},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":9.87,"kv_cache_gb":0.1,"overhead_gb":0.89,"total_gb":10.87,"total_gb_q8_cache":10.82},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":9.87,"kv_cache_gb":0.2,"overhead_gb":0.89,"total_gb":10.97,"total_gb_q8_cache":10.87},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":9.87,"kv_cache_gb":0.41,"overhead_gb":0.89,"total_gb":11.17,"total_gb_q8_cache":10.97},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":9.87,"kv_cache_gb":0.81,"overhead_gb":0.89,"total_gb":11.57,"total_gb_q8_cache":11.17},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":9.87,"kv_cache_gb":1.61,"overhead_gb":0.89,"total_gb":12.38,"total_gb_q8_cache":11.57},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":9.87,"kv_cache_gb":3.22,"overhead_gb":0.89,"total_gb":13.99,"total_gb_q8_cache":12.38},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":48,"kv_cache_gb":0.67,"overhead_gb":2.42,"total_gb":51.09,"total_gb_q8_cache":50.76},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":48,"kv_cache_gb":1.34,"overhead_gb":2.42,"total_gb":51.76,"total_gb_q8_cache":51.09},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":48,"kv_cache_gb":2.68,"overhead_gb":2.42,"total_gb":53.1,"total_gb_q8_cache":51.76},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":48,"kv_cache_gb":5.37,"overhead_gb":2.42,"total_gb":55.79,"total_gb_q8_cache":53.1},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":48,"kv_cache_gb":10.74,"overhead_gb":2.42,"total_gb":61.16,"total_gb_q8_cache":55.79},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":48,"kv_cache_gb":21.47,"overhead_gb":2.42,"total_gb":71.89,"total_gb_q8_cache":61.16},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"FP16","bytes_per_param":2,"context_tokens":262144,"weights_gb":48,"kv_cache_gb":42.95,"overhead_gb":2.42,"total_gb":93.37,"total_gb_q8_cache":71.89},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":25.44,"kv_cache_gb":0.67,"overhead_gb":1.52,"total_gb":27.63,"total_gb_q8_cache":27.29},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":25.44,"kv_cache_gb":1.34,"overhead_gb":1.52,"total_gb":28.3,"total_gb_q8_cache":27.63},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":25.44,"kv_cache_gb":2.68,"overhead_gb":1.52,"total_gb":29.64,"total_gb_q8_cache":28.3},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":25.44,"kv_cache_gb":5.37,"overhead_gb":1.52,"total_gb":32.33,"total_gb_q8_cache":29.64},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":25.44,"kv_cache_gb":10.74,"overhead_gb":1.52,"total_gb":37.7,"total_gb_q8_cache":32.33},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":25.44,"kv_cache_gb":21.47,"overhead_gb":1.52,"total_gb":48.43,"total_gb_q8_cache":37.7},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":262144,"weights_gb":25.44,"kv_cache_gb":42.95,"overhead_gb":1.52,"total_gb":69.91,"total_gb_q8_cache":48.43},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":19.68,"kv_cache_gb":0.67,"overhead_gb":1.29,"total_gb":21.64,"total_gb_q8_cache":21.3},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":19.68,"kv_cache_gb":1.34,"overhead_gb":1.29,"total_gb":22.31,"total_gb_q8_cache":21.64},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":19.68,"kv_cache_gb":2.68,"overhead_gb":1.29,"total_gb":23.65,"total_gb_q8_cache":22.31},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":19.68,"kv_cache_gb":5.37,"overhead_gb":1.29,"total_gb":26.34,"total_gb_q8_cache":23.65},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":19.68,"kv_cache_gb":10.74,"overhead_gb":1.29,"total_gb":31.7,"total_gb_q8_cache":26.34},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":19.68,"kv_cache_gb":21.47,"overhead_gb":1.29,"total_gb":42.44,"total_gb_q8_cache":31.7},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":262144,"weights_gb":19.68,"kv_cache_gb":42.95,"overhead_gb":1.29,"total_gb":63.92,"total_gb_q8_cache":42.44},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":17.04,"kv_cache_gb":0.67,"overhead_gb":1.18,"total_gb":18.89,"total_gb_q8_cache":18.56},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":17.04,"kv_cache_gb":1.34,"overhead_gb":1.18,"total_gb":19.56,"total_gb_q8_cache":18.89},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":17.04,"kv_cache_gb":2.68,"overhead_gb":1.18,"total_gb":20.91,"total_gb_q8_cache":19.56},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":17.04,"kv_cache_gb":5.37,"overhead_gb":1.18,"total_gb":23.59,"total_gb_q8_cache":20.91},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":17.04,"kv_cache_gb":10.74,"overhead_gb":1.18,"total_gb":28.96,"total_gb_q8_cache":23.59},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":17.04,"kv_cache_gb":21.47,"overhead_gb":1.18,"total_gb":39.7,"total_gb_q8_cache":28.96},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":262144,"weights_gb":17.04,"kv_cache_gb":42.95,"overhead_gb":1.18,"total_gb":61.17,"total_gb_q8_cache":39.7},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":13.92,"kv_cache_gb":0.67,"overhead_gb":1.06,"total_gb":15.65,"total_gb_q8_cache":15.31},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":13.92,"kv_cache_gb":1.34,"overhead_gb":1.06,"total_gb":16.32,"total_gb_q8_cache":15.65},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":13.92,"kv_cache_gb":2.68,"overhead_gb":1.06,"total_gb":17.66,"total_gb_q8_cache":16.32},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":13.92,"kv_cache_gb":5.37,"overhead_gb":1.06,"total_gb":20.35,"total_gb_q8_cache":17.66},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":13.92,"kv_cache_gb":10.74,"overhead_gb":1.06,"total_gb":25.71,"total_gb_q8_cache":20.35},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":13.92,"kv_cache_gb":21.47,"overhead_gb":1.06,"total_gb":36.45,"total_gb_q8_cache":25.71},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":262144,"weights_gb":13.92,"kv_cache_gb":42.95,"overhead_gb":1.06,"total_gb":57.93,"total_gb_q8_cache":36.45},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":11.28,"kv_cache_gb":0.67,"overhead_gb":0.95,"total_gb":12.9,"total_gb_q8_cache":12.57},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":11.28,"kv_cache_gb":1.34,"overhead_gb":0.95,"total_gb":13.57,"total_gb_q8_cache":12.9},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":11.28,"kv_cache_gb":2.68,"overhead_gb":0.95,"total_gb":14.92,"total_gb_q8_cache":13.57},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":11.28,"kv_cache_gb":5.37,"overhead_gb":0.95,"total_gb":17.6,"total_gb_q8_cache":14.92},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":11.28,"kv_cache_gb":10.74,"overhead_gb":0.95,"total_gb":22.97,"total_gb_q8_cache":17.6},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":11.28,"kv_cache_gb":21.47,"overhead_gb":0.95,"total_gb":33.71,"total_gb_q8_cache":22.97},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":262144,"weights_gb":11.28,"kv_cache_gb":42.95,"overhead_gb":0.95,"total_gb":55.18,"total_gb_q8_cache":33.71},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":48,"kv_cache_gb":0.67,"overhead_gb":2.42,"total_gb":51.09,"total_gb_q8_cache":50.76},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":48,"kv_cache_gb":1.34,"overhead_gb":2.42,"total_gb":51.76,"total_gb_q8_cache":51.09},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":48,"kv_cache_gb":2.68,"overhead_gb":2.42,"total_gb":53.1,"total_gb_q8_cache":51.76},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":48,"kv_cache_gb":5.37,"overhead_gb":2.42,"total_gb":55.79,"total_gb_q8_cache":53.1},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":48,"kv_cache_gb":10.74,"overhead_gb":2.42,"total_gb":61.16,"total_gb_q8_cache":55.79},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":48,"kv_cache_gb":21.47,"overhead_gb":2.42,"total_gb":71.89,"total_gb_q8_cache":61.16},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":25.44,"kv_cache_gb":0.67,"overhead_gb":1.52,"total_gb":27.63,"total_gb_q8_cache":27.29},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":25.44,"kv_cache_gb":1.34,"overhead_gb":1.52,"total_gb":28.3,"total_gb_q8_cache":27.63},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":25.44,"kv_cache_gb":2.68,"overhead_gb":1.52,"total_gb":29.64,"total_gb_q8_cache":28.3},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":25.44,"kv_cache_gb":5.37,"overhead_gb":1.52,"total_gb":32.33,"total_gb_q8_cache":29.64},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":25.44,"kv_cache_gb":10.74,"overhead_gb":1.52,"total_gb":37.7,"total_gb_q8_cache":32.33},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":25.44,"kv_cache_gb":21.47,"overhead_gb":1.52,"total_gb":48.43,"total_gb_q8_cache":37.7},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":19.68,"kv_cache_gb":0.67,"overhead_gb":1.29,"total_gb":21.64,"total_gb_q8_cache":21.3},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":19.68,"kv_cache_gb":1.34,"overhead_gb":1.29,"total_gb":22.31,"total_gb_q8_cache":21.64},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":19.68,"kv_cache_gb":2.68,"overhead_gb":1.29,"total_gb":23.65,"total_gb_q8_cache":22.31},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":19.68,"kv_cache_gb":5.37,"overhead_gb":1.29,"total_gb":26.34,"total_gb_q8_cache":23.65},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":19.68,"kv_cache_gb":10.74,"overhead_gb":1.29,"total_gb":31.7,"total_gb_q8_cache":26.34},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":19.68,"kv_cache_gb":21.47,"overhead_gb":1.29,"total_gb":42.44,"total_gb_q8_cache":31.7},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":17.04,"kv_cache_gb":0.67,"overhead_gb":1.18,"total_gb":18.89,"total_gb_q8_cache":18.56},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":17.04,"kv_cache_gb":1.34,"overhead_gb":1.18,"total_gb":19.56,"total_gb_q8_cache":18.89},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":17.04,"kv_cache_gb":2.68,"overhead_gb":1.18,"total_gb":20.91,"total_gb_q8_cache":19.56},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":17.04,"kv_cache_gb":5.37,"overhead_gb":1.18,"total_gb":23.59,"total_gb_q8_cache":20.91},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":17.04,"kv_cache_gb":10.74,"overhead_gb":1.18,"total_gb":28.96,"total_gb_q8_cache":23.59},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":17.04,"kv_cache_gb":21.47,"overhead_gb":1.18,"total_gb":39.7,"total_gb_q8_cache":28.96},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":13.92,"kv_cache_gb":0.67,"overhead_gb":1.06,"total_gb":15.65,"total_gb_q8_cache":15.31},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":13.92,"kv_cache_gb":1.34,"overhead_gb":1.06,"total_gb":16.32,"total_gb_q8_cache":15.65},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":13.92,"kv_cache_gb":2.68,"overhead_gb":1.06,"total_gb":17.66,"total_gb_q8_cache":16.32},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":13.92,"kv_cache_gb":5.37,"overhead_gb":1.06,"total_gb":20.35,"total_gb_q8_cache":17.66},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":13.92,"kv_cache_gb":10.74,"overhead_gb":1.06,"total_gb":25.71,"total_gb_q8_cache":20.35},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":13.92,"kv_cache_gb":21.47,"overhead_gb":1.06,"total_gb":36.45,"total_gb_q8_cache":25.71},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":11.28,"kv_cache_gb":0.67,"overhead_gb":0.95,"total_gb":12.9,"total_gb_q8_cache":12.57},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":11.28,"kv_cache_gb":1.34,"overhead_gb":0.95,"total_gb":13.57,"total_gb_q8_cache":12.9},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":11.28,"kv_cache_gb":2.68,"overhead_gb":0.95,"total_gb":14.92,"total_gb_q8_cache":13.57},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":11.28,"kv_cache_gb":5.37,"overhead_gb":0.95,"total_gb":17.6,"total_gb_q8_cache":14.92},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":11.28,"kv_cache_gb":10.74,"overhead_gb":0.95,"total_gb":22.97,"total_gb_q8_cache":17.6},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":11.28,"kv_cache_gb":21.47,"overhead_gb":0.95,"total_gb":33.71,"total_gb_q8_cache":22.97},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":51.6,"kv_cache_gb":0.29,"overhead_gb":2.56,"total_gb":54.46,"total_gb_q8_cache":54.31},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":51.6,"kv_cache_gb":0.38,"overhead_gb":2.56,"total_gb":54.54,"total_gb_q8_cache":54.35},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":51.6,"kv_cache_gb":0.55,"overhead_gb":2.56,"total_gb":54.71,"total_gb_q8_cache":54.44},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":51.6,"kv_cache_gb":0.88,"overhead_gb":2.56,"total_gb":55.04,"total_gb_q8_cache":54.6},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":51.6,"kv_cache_gb":1.55,"overhead_gb":2.56,"total_gb":55.72,"total_gb_q8_cache":54.94},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":51.6,"kv_cache_gb":2.89,"overhead_gb":2.56,"total_gb":57.06,"total_gb_q8_cache":55.61},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":262144,"weights_gb":51.6,"kv_cache_gb":5.58,"overhead_gb":2.56,"total_gb":59.74,"total_gb_q8_cache":56.95},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":27.35,"kv_cache_gb":0.29,"overhead_gb":1.59,"total_gb":29.24,"total_gb_q8_cache":29.09},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":27.35,"kv_cache_gb":0.38,"overhead_gb":1.59,"total_gb":29.32,"total_gb_q8_cache":29.13},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":27.35,"kv_cache_gb":0.55,"overhead_gb":1.59,"total_gb":29.49,"total_gb_q8_cache":29.21},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":27.35,"kv_cache_gb":0.88,"overhead_gb":1.59,"total_gb":29.82,"total_gb_q8_cache":29.38},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":27.35,"kv_cache_gb":1.55,"overhead_gb":1.59,"total_gb":30.49,"total_gb_q8_cache":29.72},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":27.35,"kv_cache_gb":2.89,"overhead_gb":1.59,"total_gb":31.84,"total_gb_q8_cache":30.39},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":262144,"weights_gb":27.35,"kv_cache_gb":5.58,"overhead_gb":1.59,"total_gb":34.52,"total_gb_q8_cache":31.73},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":21.16,"kv_cache_gb":0.29,"overhead_gb":1.35,"total_gb":22.8,"total_gb_q8_cache":22.65},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":21.16,"kv_cache_gb":0.38,"overhead_gb":1.35,"total_gb":22.88,"total_gb_q8_cache":22.69},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":21.16,"kv_cache_gb":0.55,"overhead_gb":1.35,"total_gb":23.05,"total_gb_q8_cache":22.77},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":21.16,"kv_cache_gb":0.88,"overhead_gb":1.35,"total_gb":23.38,"total_gb_q8_cache":22.94},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":21.16,"kv_cache_gb":1.55,"overhead_gb":1.35,"total_gb":24.05,"total_gb_q8_cache":23.28},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":21.16,"kv_cache_gb":2.89,"overhead_gb":1.35,"total_gb":25.4,"total_gb_q8_cache":23.95},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":262144,"weights_gb":21.16,"kv_cache_gb":5.58,"overhead_gb":1.35,"total_gb":28.08,"total_gb_q8_cache":25.29},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":18.32,"kv_cache_gb":0.29,"overhead_gb":1.23,"total_gb":19.84,"total_gb_q8_cache":19.7},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":18.32,"kv_cache_gb":0.38,"overhead_gb":1.23,"total_gb":19.93,"total_gb_q8_cache":19.74},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":18.32,"kv_cache_gb":0.55,"overhead_gb":1.23,"total_gb":20.1,"total_gb_q8_cache":19.82},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":18.32,"kv_cache_gb":0.88,"overhead_gb":1.23,"total_gb":20.43,"total_gb_q8_cache":19.99},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":18.32,"kv_cache_gb":1.55,"overhead_gb":1.23,"total_gb":21.1,"total_gb_q8_cache":20.33},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":18.32,"kv_cache_gb":2.89,"overhead_gb":1.23,"total_gb":22.44,"total_gb_q8_cache":21},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":262144,"weights_gb":18.32,"kv_cache_gb":5.58,"overhead_gb":1.23,"total_gb":25.13,"total_gb_q8_cache":22.34},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":14.96,"kv_cache_gb":0.29,"overhead_gb":1.1,"total_gb":16.36,"total_gb_q8_cache":16.21},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":14.96,"kv_cache_gb":0.38,"overhead_gb":1.1,"total_gb":16.44,"total_gb_q8_cache":16.25},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":14.96,"kv_cache_gb":0.55,"overhead_gb":1.1,"total_gb":16.61,"total_gb_q8_cache":16.34},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":14.96,"kv_cache_gb":0.88,"overhead_gb":1.1,"total_gb":16.94,"total_gb_q8_cache":16.5},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":14.96,"kv_cache_gb":1.55,"overhead_gb":1.1,"total_gb":17.61,"total_gb_q8_cache":16.84},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":14.96,"kv_cache_gb":2.89,"overhead_gb":1.1,"total_gb":18.96,"total_gb_q8_cache":17.51},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":262144,"weights_gb":14.96,"kv_cache_gb":5.58,"overhead_gb":1.1,"total_gb":21.64,"total_gb_q8_cache":18.85},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":12.13,"kv_cache_gb":0.29,"overhead_gb":0.99,"total_gb":13.4,"total_gb_q8_cache":13.26},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":12.13,"kv_cache_gb":0.38,"overhead_gb":0.99,"total_gb":13.49,"total_gb_q8_cache":13.3},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":12.13,"kv_cache_gb":0.55,"overhead_gb":0.99,"total_gb":13.66,"total_gb_q8_cache":13.38},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":12.13,"kv_cache_gb":0.88,"overhead_gb":0.99,"total_gb":13.99,"total_gb_q8_cache":13.55},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":12.13,"kv_cache_gb":1.55,"overhead_gb":0.99,"total_gb":14.66,"total_gb_q8_cache":13.89},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":12.13,"kv_cache_gb":2.89,"overhead_gb":0.99,"total_gb":16.01,"total_gb_q8_cache":14.56},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":262144,"weights_gb":12.13,"kv_cache_gb":5.58,"overhead_gb":0.99,"total_gb":18.69,"total_gb_q8_cache":15.9},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":54.8,"kv_cache_gb":0.77,"overhead_gb":2.69,"total_gb":58.26,"total_gb_q8_cache":57.88},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":54.8,"kv_cache_gb":1.11,"overhead_gb":2.69,"total_gb":58.6,"total_gb_q8_cache":58.05},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":54.8,"kv_cache_gb":1.78,"overhead_gb":2.69,"total_gb":59.27,"total_gb_q8_cache":58.38},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":54.8,"kv_cache_gb":3.12,"overhead_gb":2.69,"total_gb":60.61,"total_gb_q8_cache":59.05},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":54.8,"kv_cache_gb":5.8,"overhead_gb":2.69,"total_gb":63.3,"total_gb_q8_cache":60.39},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":54.8,"kv_cache_gb":11.17,"overhead_gb":2.69,"total_gb":68.67,"total_gb_q8_cache":63.08},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":29.04,"kv_cache_gb":0.77,"overhead_gb":1.66,"total_gb":31.48,"total_gb_q8_cache":31.09},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":29.04,"kv_cache_gb":1.11,"overhead_gb":1.66,"total_gb":31.81,"total_gb_q8_cache":31.26},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":29.04,"kv_cache_gb":1.78,"overhead_gb":1.66,"total_gb":32.48,"total_gb_q8_cache":31.59},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":29.04,"kv_cache_gb":3.12,"overhead_gb":1.66,"total_gb":33.83,"total_gb_q8_cache":32.27},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":29.04,"kv_cache_gb":5.8,"overhead_gb":1.66,"total_gb":36.51,"total_gb_q8_cache":33.61},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":29.04,"kv_cache_gb":11.17,"overhead_gb":1.66,"total_gb":41.88,"total_gb_q8_cache":36.29},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":22.47,"kv_cache_gb":0.77,"overhead_gb":1.4,"total_gb":24.64,"total_gb_q8_cache":24.25},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":22.47,"kv_cache_gb":1.11,"overhead_gb":1.4,"total_gb":24.97,"total_gb_q8_cache":24.42},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":22.47,"kv_cache_gb":1.78,"overhead_gb":1.4,"total_gb":25.65,"total_gb_q8_cache":24.76},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":22.47,"kv_cache_gb":3.12,"overhead_gb":1.4,"total_gb":26.99,"total_gb_q8_cache":25.43},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":22.47,"kv_cache_gb":5.8,"overhead_gb":1.4,"total_gb":29.67,"total_gb_q8_cache":26.77},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":22.47,"kv_cache_gb":11.17,"overhead_gb":1.4,"total_gb":35.04,"total_gb_q8_cache":29.45},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":19.45,"kv_cache_gb":0.77,"overhead_gb":1.28,"total_gb":21.5,"total_gb_q8_cache":21.12},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":19.45,"kv_cache_gb":1.11,"overhead_gb":1.28,"total_gb":21.84,"total_gb_q8_cache":21.29},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":19.45,"kv_cache_gb":1.78,"overhead_gb":1.28,"total_gb":22.51,"total_gb_q8_cache":21.62},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":19.45,"kv_cache_gb":3.12,"overhead_gb":1.28,"total_gb":23.85,"total_gb_q8_cache":22.29},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":19.45,"kv_cache_gb":5.8,"overhead_gb":1.28,"total_gb":26.54,"total_gb_q8_cache":23.63},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":19.45,"kv_cache_gb":11.17,"overhead_gb":1.28,"total_gb":31.91,"total_gb_q8_cache":26.32},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":15.89,"kv_cache_gb":0.77,"overhead_gb":1.14,"total_gb":17.8,"total_gb_q8_cache":17.41},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":15.89,"kv_cache_gb":1.11,"overhead_gb":1.14,"total_gb":18.13,"total_gb_q8_cache":17.58},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":15.89,"kv_cache_gb":1.78,"overhead_gb":1.14,"total_gb":18.81,"total_gb_q8_cache":17.92},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":15.89,"kv_cache_gb":3.12,"overhead_gb":1.14,"total_gb":20.15,"total_gb_q8_cache":18.59},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":15.89,"kv_cache_gb":5.8,"overhead_gb":1.14,"total_gb":22.83,"total_gb_q8_cache":19.93},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":15.89,"kv_cache_gb":11.17,"overhead_gb":1.14,"total_gb":28.2,"total_gb_q8_cache":22.61},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":12.88,"kv_cache_gb":0.77,"overhead_gb":1.02,"total_gb":14.66,"total_gb_q8_cache":14.28},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":12.88,"kv_cache_gb":1.11,"overhead_gb":1.02,"total_gb":15,"total_gb_q8_cache":14.45},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":12.88,"kv_cache_gb":1.78,"overhead_gb":1.02,"total_gb":15.67,"total_gb_q8_cache":14.78},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":12.88,"kv_cache_gb":3.12,"overhead_gb":1.02,"total_gb":17.01,"total_gb_q8_cache":15.45},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":12.88,"kv_cache_gb":5.8,"overhead_gb":1.02,"total_gb":19.7,"total_gb_q8_cache":16.8},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":12.88,"kv_cache_gb":11.17,"overhead_gb":1.02,"total_gb":25.07,"total_gb_q8_cache":19.48},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":55.6,"kv_cache_gb":0.42,"overhead_gb":2.72,"total_gb":58.74,"total_gb_q8_cache":58.61},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":55.6,"kv_cache_gb":0.69,"overhead_gb":2.72,"total_gb":59.01,"total_gb_q8_cache":58.74},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":55.6,"kv_cache_gb":1.22,"overhead_gb":2.72,"total_gb":59.55,"total_gb_q8_cache":59.01},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":55.6,"kv_cache_gb":2.3,"overhead_gb":2.72,"total_gb":60.62,"total_gb_q8_cache":59.55},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":55.6,"kv_cache_gb":4.44,"overhead_gb":2.72,"total_gb":62.77,"total_gb_q8_cache":60.62},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":55.6,"kv_cache_gb":8.74,"overhead_gb":2.72,"total_gb":67.06,"total_gb_q8_cache":62.77},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"FP16","bytes_per_param":2,"context_tokens":262144,"weights_gb":55.6,"kv_cache_gb":17.33,"overhead_gb":2.72,"total_gb":75.65,"total_gb_q8_cache":67.06},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":29.47,"kv_cache_gb":0.42,"overhead_gb":1.68,"total_gb":31.57,"total_gb_q8_cache":31.43},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":29.47,"kv_cache_gb":0.69,"overhead_gb":1.68,"total_gb":31.83,"total_gb_q8_cache":31.57},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":29.47,"kv_cache_gb":1.22,"overhead_gb":1.68,"total_gb":32.37,"total_gb_q8_cache":31.83},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":29.47,"kv_cache_gb":2.3,"overhead_gb":1.68,"total_gb":33.44,"total_gb_q8_cache":32.37},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":29.47,"kv_cache_gb":4.44,"overhead_gb":1.68,"total_gb":35.59,"total_gb_q8_cache":33.44},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":29.47,"kv_cache_gb":8.74,"overhead_gb":1.68,"total_gb":39.89,"total_gb_q8_cache":35.59},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":262144,"weights_gb":29.47,"kv_cache_gb":17.33,"overhead_gb":1.68,"total_gb":48.48,"total_gb_q8_cache":39.89},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":22.8,"kv_cache_gb":0.42,"overhead_gb":1.41,"total_gb":24.63,"total_gb_q8_cache":24.49},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":22.8,"kv_cache_gb":0.69,"overhead_gb":1.41,"total_gb":24.89,"total_gb_q8_cache":24.63},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":22.8,"kv_cache_gb":1.22,"overhead_gb":1.41,"total_gb":25.43,"total_gb_q8_cache":24.89},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":22.8,"kv_cache_gb":2.3,"overhead_gb":1.41,"total_gb":26.51,"total_gb_q8_cache":25.43},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":22.8,"kv_cache_gb":4.44,"overhead_gb":1.41,"total_gb":28.65,"total_gb_q8_cache":26.51},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":22.8,"kv_cache_gb":8.74,"overhead_gb":1.41,"total_gb":32.95,"total_gb_q8_cache":28.65},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":262144,"weights_gb":22.8,"kv_cache_gb":17.33,"overhead_gb":1.41,"total_gb":41.54,"total_gb_q8_cache":32.95},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":19.74,"kv_cache_gb":0.42,"overhead_gb":1.29,"total_gb":21.45,"total_gb_q8_cache":21.31},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":19.74,"kv_cache_gb":0.69,"overhead_gb":1.29,"total_gb":21.71,"total_gb_q8_cache":21.45},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":19.74,"kv_cache_gb":1.22,"overhead_gb":1.29,"total_gb":22.25,"total_gb_q8_cache":21.71},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":19.74,"kv_cache_gb":2.3,"overhead_gb":1.29,"total_gb":23.33,"total_gb_q8_cache":22.25},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":19.74,"kv_cache_gb":4.44,"overhead_gb":1.29,"total_gb":25.47,"total_gb_q8_cache":23.33},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":19.74,"kv_cache_gb":8.74,"overhead_gb":1.29,"total_gb":29.77,"total_gb_q8_cache":25.47},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":262144,"weights_gb":19.74,"kv_cache_gb":17.33,"overhead_gb":1.29,"total_gb":38.36,"total_gb_q8_cache":29.77},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":16.12,"kv_cache_gb":0.42,"overhead_gb":1.14,"total_gb":17.69,"total_gb_q8_cache":17.55},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":16.12,"kv_cache_gb":0.69,"overhead_gb":1.14,"total_gb":17.96,"total_gb_q8_cache":17.69},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":16.12,"kv_cache_gb":1.22,"overhead_gb":1.14,"total_gb":18.49,"total_gb_q8_cache":17.96},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":16.12,"kv_cache_gb":2.3,"overhead_gb":1.14,"total_gb":19.57,"total_gb_q8_cache":18.49},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":16.12,"kv_cache_gb":4.44,"overhead_gb":1.14,"total_gb":21.71,"total_gb_q8_cache":19.57},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":16.12,"kv_cache_gb":8.74,"overhead_gb":1.14,"total_gb":26.01,"total_gb_q8_cache":21.71},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":262144,"weights_gb":16.12,"kv_cache_gb":17.33,"overhead_gb":1.14,"total_gb":34.6,"total_gb_q8_cache":26.01},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":13.07,"kv_cache_gb":0.42,"overhead_gb":1.02,"total_gb":14.51,"total_gb_q8_cache":14.37},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":13.07,"kv_cache_gb":0.69,"overhead_gb":1.02,"total_gb":14.78,"total_gb_q8_cache":14.51},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":13.07,"kv_cache_gb":1.22,"overhead_gb":1.02,"total_gb":15.31,"total_gb_q8_cache":14.78},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":13.07,"kv_cache_gb":2.3,"overhead_gb":1.02,"total_gb":16.39,"total_gb_q8_cache":15.31},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":13.07,"kv_cache_gb":4.44,"overhead_gb":1.02,"total_gb":18.53,"total_gb_q8_cache":16.39},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":13.07,"kv_cache_gb":8.74,"overhead_gb":1.02,"total_gb":22.83,"total_gb_q8_cache":18.53},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":262144,"weights_gb":13.07,"kv_cache_gb":17.33,"overhead_gb":1.02,"total_gb":31.42,"total_gb_q8_cache":22.83},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":61,"kv_cache_gb":0.4,"overhead_gb":2.94,"total_gb":64.34,"total_gb_q8_cache":64.14},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":61,"kv_cache_gb":0.81,"overhead_gb":2.94,"total_gb":64.75,"total_gb_q8_cache":64.34},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":61,"kv_cache_gb":1.61,"overhead_gb":2.94,"total_gb":65.55,"total_gb_q8_cache":64.75},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":61,"kv_cache_gb":3.22,"overhead_gb":2.94,"total_gb":67.16,"total_gb_q8_cache":65.55},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":40960,"weights_gb":61,"kv_cache_gb":4.03,"overhead_gb":2.94,"total_gb":67.97,"total_gb_q8_cache":65.95},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":32.33,"kv_cache_gb":0.4,"overhead_gb":1.79,"total_gb":34.53,"total_gb_q8_cache":34.32},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":32.33,"kv_cache_gb":0.81,"overhead_gb":1.79,"total_gb":34.93,"total_gb_q8_cache":34.53},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":32.33,"kv_cache_gb":1.61,"overhead_gb":1.79,"total_gb":35.73,"total_gb_q8_cache":34.93},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":32.33,"kv_cache_gb":3.22,"overhead_gb":1.79,"total_gb":37.34,"total_gb_q8_cache":35.73},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":40960,"weights_gb":32.33,"kv_cache_gb":4.03,"overhead_gb":1.79,"total_gb":38.15,"total_gb_q8_cache":36.14},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":25.01,"kv_cache_gb":0.4,"overhead_gb":1.5,"total_gb":26.91,"total_gb_q8_cache":26.71},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":25.01,"kv_cache_gb":0.81,"overhead_gb":1.5,"total_gb":27.32,"total_gb_q8_cache":26.91},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":25.01,"kv_cache_gb":1.61,"overhead_gb":1.5,"total_gb":28.12,"total_gb_q8_cache":27.32},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":25.01,"kv_cache_gb":3.22,"overhead_gb":1.5,"total_gb":29.73,"total_gb_q8_cache":28.12},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":40960,"weights_gb":25.01,"kv_cache_gb":4.03,"overhead_gb":1.5,"total_gb":30.54,"total_gb_q8_cache":28.52},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":21.65,"kv_cache_gb":0.4,"overhead_gb":1.37,"total_gb":23.42,"total_gb_q8_cache":23.22},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":21.65,"kv_cache_gb":0.81,"overhead_gb":1.37,"total_gb":23.83,"total_gb_q8_cache":23.42},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":21.65,"kv_cache_gb":1.61,"overhead_gb":1.37,"total_gb":24.63,"total_gb_q8_cache":23.83},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":21.65,"kv_cache_gb":3.22,"overhead_gb":1.37,"total_gb":26.24,"total_gb_q8_cache":24.63},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":40960,"weights_gb":21.65,"kv_cache_gb":4.03,"overhead_gb":1.37,"total_gb":27.05,"total_gb_q8_cache":25.03},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":17.69,"kv_cache_gb":0.4,"overhead_gb":1.21,"total_gb":19.3,"total_gb_q8_cache":19.1},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":17.69,"kv_cache_gb":0.81,"overhead_gb":1.21,"total_gb":19.7,"total_gb_q8_cache":19.3},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":17.69,"kv_cache_gb":1.61,"overhead_gb":1.21,"total_gb":20.51,"total_gb_q8_cache":19.7},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":17.69,"kv_cache_gb":3.22,"overhead_gb":1.21,"total_gb":22.12,"total_gb_q8_cache":20.51},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":40960,"weights_gb":17.69,"kv_cache_gb":4.03,"overhead_gb":1.21,"total_gb":22.92,"total_gb_q8_cache":20.91},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":14.34,"kv_cache_gb":0.4,"overhead_gb":1.07,"total_gb":15.81,"total_gb_q8_cache":15.61},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":14.34,"kv_cache_gb":0.81,"overhead_gb":1.07,"total_gb":16.21,"total_gb_q8_cache":15.81},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":14.34,"kv_cache_gb":1.61,"overhead_gb":1.07,"total_gb":17.02,"total_gb_q8_cache":16.21},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":14.34,"kv_cache_gb":3.22,"overhead_gb":1.07,"total_gb":18.63,"total_gb_q8_cache":17.02},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":40960,"weights_gb":14.34,"kv_cache_gb":4.03,"overhead_gb":1.07,"total_gb":19.43,"total_gb_q8_cache":17.42},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":62.4,"kv_cache_gb":0.22,"overhead_gb":3,"total_gb":65.62,"total_gb_q8_cache":65.51},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":62.4,"kv_cache_gb":0.44,"overhead_gb":3,"total_gb":65.84,"total_gb_q8_cache":65.62},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":62.4,"kv_cache_gb":0.89,"overhead_gb":3,"total_gb":66.28,"total_gb_q8_cache":65.84},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":62.4,"kv_cache_gb":1.77,"overhead_gb":3,"total_gb":67.17,"total_gb_q8_cache":66.28},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":62.4,"kv_cache_gb":3.55,"overhead_gb":3,"total_gb":68.94,"total_gb_q8_cache":67.17},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":62.4,"kv_cache_gb":7.1,"overhead_gb":3,"total_gb":72.49,"total_gb_q8_cache":68.94},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":202752,"weights_gb":62.4,"kv_cache_gb":10.98,"overhead_gb":3,"total_gb":76.37,"total_gb_q8_cache":70.88},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":33.07,"kv_cache_gb":0.22,"overhead_gb":1.82,"total_gb":35.12,"total_gb_q8_cache":35.01},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":33.07,"kv_cache_gb":0.44,"overhead_gb":1.82,"total_gb":35.34,"total_gb_q8_cache":35.12},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":33.07,"kv_cache_gb":0.89,"overhead_gb":1.82,"total_gb":35.78,"total_gb_q8_cache":35.34},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":33.07,"kv_cache_gb":1.77,"overhead_gb":1.82,"total_gb":36.67,"total_gb_q8_cache":35.78},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":33.07,"kv_cache_gb":3.55,"overhead_gb":1.82,"total_gb":38.44,"total_gb_q8_cache":36.67},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":33.07,"kv_cache_gb":7.1,"overhead_gb":1.82,"total_gb":41.99,"total_gb_q8_cache":38.44},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":202752,"weights_gb":33.07,"kv_cache_gb":10.98,"overhead_gb":1.82,"total_gb":45.87,"total_gb_q8_cache":40.38},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":25.58,"kv_cache_gb":0.22,"overhead_gb":1.52,"total_gb":27.33,"total_gb_q8_cache":27.22},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":25.58,"kv_cache_gb":0.44,"overhead_gb":1.52,"total_gb":27.55,"total_gb_q8_cache":27.33},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":25.58,"kv_cache_gb":0.89,"overhead_gb":1.52,"total_gb":27.99,"total_gb_q8_cache":27.55},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":25.58,"kv_cache_gb":1.77,"overhead_gb":1.52,"total_gb":28.88,"total_gb_q8_cache":27.99},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":25.58,"kv_cache_gb":3.55,"overhead_gb":1.52,"total_gb":30.66,"total_gb_q8_cache":28.88},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":25.58,"kv_cache_gb":7.1,"overhead_gb":1.52,"total_gb":34.2,"total_gb_q8_cache":30.66},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":202752,"weights_gb":25.58,"kv_cache_gb":10.98,"overhead_gb":1.52,"total_gb":38.09,"total_gb_q8_cache":32.6},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":22.15,"kv_cache_gb":0.22,"overhead_gb":1.39,"total_gb":23.76,"total_gb_q8_cache":23.65},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":22.15,"kv_cache_gb":0.44,"overhead_gb":1.39,"total_gb":23.98,"total_gb_q8_cache":23.76},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":22.15,"kv_cache_gb":0.89,"overhead_gb":1.39,"total_gb":24.43,"total_gb_q8_cache":23.98},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":22.15,"kv_cache_gb":1.77,"overhead_gb":1.39,"total_gb":25.31,"total_gb_q8_cache":24.43},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":22.15,"kv_cache_gb":3.55,"overhead_gb":1.39,"total_gb":27.09,"total_gb_q8_cache":25.31},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":22.15,"kv_cache_gb":7.1,"overhead_gb":1.39,"total_gb":30.63,"total_gb_q8_cache":27.09},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":202752,"weights_gb":22.15,"kv_cache_gb":10.98,"overhead_gb":1.39,"total_gb":34.52,"total_gb_q8_cache":29.03},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":18.1,"kv_cache_gb":0.22,"overhead_gb":1.22,"total_gb":19.54,"total_gb_q8_cache":19.43},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":18.1,"kv_cache_gb":0.44,"overhead_gb":1.22,"total_gb":19.76,"total_gb_q8_cache":19.54},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":18.1,"kv_cache_gb":0.89,"overhead_gb":1.22,"total_gb":20.21,"total_gb_q8_cache":19.76},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":18.1,"kv_cache_gb":1.77,"overhead_gb":1.22,"total_gb":21.09,"total_gb_q8_cache":20.21},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":18.1,"kv_cache_gb":3.55,"overhead_gb":1.22,"total_gb":22.87,"total_gb_q8_cache":21.09},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":18.1,"kv_cache_gb":7.1,"overhead_gb":1.22,"total_gb":26.42,"total_gb_q8_cache":22.87},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":202752,"weights_gb":18.1,"kv_cache_gb":10.98,"overhead_gb":1.22,"total_gb":30.3,"total_gb_q8_cache":24.81},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":14.66,"kv_cache_gb":0.22,"overhead_gb":1.09,"total_gb":15.97,"total_gb_q8_cache":15.86},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":14.66,"kv_cache_gb":0.44,"overhead_gb":1.09,"total_gb":16.19,"total_gb_q8_cache":15.97},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":14.66,"kv_cache_gb":0.89,"overhead_gb":1.09,"total_gb":16.64,"total_gb_q8_cache":16.19},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":14.66,"kv_cache_gb":1.77,"overhead_gb":1.09,"total_gb":17.52,"total_gb_q8_cache":16.64},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":14.66,"kv_cache_gb":3.55,"overhead_gb":1.09,"total_gb":19.3,"total_gb_q8_cache":17.52},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":14.66,"kv_cache_gb":7.1,"overhead_gb":1.09,"total_gb":22.85,"total_gb_q8_cache":19.3},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":202752,"weights_gb":14.66,"kv_cache_gb":10.98,"overhead_gb":1.09,"total_gb":26.73,"total_gb_q8_cache":21.24},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":62.6,"kv_cache_gb":1.17,"overhead_gb":3,"total_gb":66.78,"total_gb_q8_cache":66.19},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":62.6,"kv_cache_gb":1.51,"overhead_gb":3,"total_gb":67.11,"total_gb_q8_cache":66.36},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":62.6,"kv_cache_gb":2.18,"overhead_gb":3,"total_gb":67.79,"total_gb_q8_cache":66.69},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":62.6,"kv_cache_gb":3.52,"overhead_gb":3,"total_gb":69.13,"total_gb_q8_cache":67.37},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":62.6,"kv_cache_gb":6.21,"overhead_gb":3,"total_gb":71.81,"total_gb_q8_cache":68.71},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":62.6,"kv_cache_gb":11.58,"overhead_gb":3,"total_gb":77.18,"total_gb_q8_cache":71.39},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"FP16","bytes_per_param":2,"context_tokens":262144,"weights_gb":62.6,"kv_cache_gb":22.31,"overhead_gb":3,"total_gb":87.92,"total_gb_q8_cache":76.76},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":33.18,"kv_cache_gb":1.17,"overhead_gb":1.83,"total_gb":36.18,"total_gb_q8_cache":35.59},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":33.18,"kv_cache_gb":1.51,"overhead_gb":1.83,"total_gb":36.52,"total_gb_q8_cache":35.76},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":33.18,"kv_cache_gb":2.18,"overhead_gb":1.83,"total_gb":37.19,"total_gb_q8_cache":36.1},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":33.18,"kv_cache_gb":3.52,"overhead_gb":1.83,"total_gb":38.53,"total_gb_q8_cache":36.77},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":33.18,"kv_cache_gb":6.21,"overhead_gb":1.83,"total_gb":41.21,"total_gb_q8_cache":38.11},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":33.18,"kv_cache_gb":11.58,"overhead_gb":1.83,"total_gb":46.58,"total_gb_q8_cache":40.79},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":262144,"weights_gb":33.18,"kv_cache_gb":22.31,"overhead_gb":1.83,"total_gb":57.32,"total_gb_q8_cache":46.16},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":25.67,"kv_cache_gb":1.17,"overhead_gb":1.53,"total_gb":28.37,"total_gb_q8_cache":27.78},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":25.67,"kv_cache_gb":1.51,"overhead_gb":1.53,"total_gb":28.7,"total_gb_q8_cache":27.95},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":25.67,"kv_cache_gb":2.18,"overhead_gb":1.53,"total_gb":29.37,"total_gb_q8_cache":28.28},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":25.67,"kv_cache_gb":3.52,"overhead_gb":1.53,"total_gb":30.72,"total_gb_q8_cache":28.95},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":25.67,"kv_cache_gb":6.21,"overhead_gb":1.53,"total_gb":33.4,"total_gb_q8_cache":30.3},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":25.67,"kv_cache_gb":11.58,"overhead_gb":1.53,"total_gb":38.77,"total_gb_q8_cache":32.98},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":262144,"weights_gb":25.67,"kv_cache_gb":22.31,"overhead_gb":1.53,"total_gb":49.51,"total_gb_q8_cache":38.35},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":22.22,"kv_cache_gb":1.17,"overhead_gb":1.39,"total_gb":24.79,"total_gb_q8_cache":24.2},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":22.22,"kv_cache_gb":1.51,"overhead_gb":1.39,"total_gb":25.12,"total_gb_q8_cache":24.37},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":22.22,"kv_cache_gb":2.18,"overhead_gb":1.39,"total_gb":25.79,"total_gb_q8_cache":24.7},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":22.22,"kv_cache_gb":3.52,"overhead_gb":1.39,"total_gb":27.14,"total_gb_q8_cache":25.37},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":22.22,"kv_cache_gb":6.21,"overhead_gb":1.39,"total_gb":29.82,"total_gb_q8_cache":26.72},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":22.22,"kv_cache_gb":11.58,"overhead_gb":1.39,"total_gb":35.19,"total_gb_q8_cache":29.4},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":262144,"weights_gb":22.22,"kv_cache_gb":22.31,"overhead_gb":1.39,"total_gb":45.93,"total_gb_q8_cache":34.77},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":18.15,"kv_cache_gb":1.17,"overhead_gb":1.23,"total_gb":20.55,"total_gb_q8_cache":19.97},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":18.15,"kv_cache_gb":1.51,"overhead_gb":1.23,"total_gb":20.89,"total_gb_q8_cache":20.14},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":18.15,"kv_cache_gb":2.18,"overhead_gb":1.23,"total_gb":21.56,"total_gb_q8_cache":20.47},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":18.15,"kv_cache_gb":3.52,"overhead_gb":1.23,"total_gb":22.9,"total_gb_q8_cache":21.14},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":18.15,"kv_cache_gb":6.21,"overhead_gb":1.23,"total_gb":25.59,"total_gb_q8_cache":22.48},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":18.15,"kv_cache_gb":11.58,"overhead_gb":1.23,"total_gb":30.96,"total_gb_q8_cache":25.17},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":262144,"weights_gb":18.15,"kv_cache_gb":22.31,"overhead_gb":1.23,"total_gb":41.69,"total_gb_q8_cache":30.54},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":14.71,"kv_cache_gb":1.17,"overhead_gb":1.09,"total_gb":16.97,"total_gb_q8_cache":16.39},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":14.71,"kv_cache_gb":1.51,"overhead_gb":1.09,"total_gb":17.31,"total_gb_q8_cache":16.55},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":14.71,"kv_cache_gb":2.18,"overhead_gb":1.09,"total_gb":17.98,"total_gb_q8_cache":16.89},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":14.71,"kv_cache_gb":3.52,"overhead_gb":1.09,"total_gb":19.32,"total_gb_q8_cache":17.56},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":14.71,"kv_cache_gb":6.21,"overhead_gb":1.09,"total_gb":22.01,"total_gb_q8_cache":18.9},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":14.71,"kv_cache_gb":11.58,"overhead_gb":1.09,"total_gb":27.38,"total_gb_q8_cache":21.59},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":262144,"weights_gb":14.71,"kv_cache_gb":22.31,"overhead_gb":1.09,"total_gb":38.11,"total_gb_q8_cache":26.96},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":63.2,"kv_cache_gb":0.08,"overhead_gb":3.03,"total_gb":66.3,"total_gb_q8_cache":66.29},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":63.2,"kv_cache_gb":0.1,"overhead_gb":3.03,"total_gb":66.33,"total_gb_q8_cache":66.3},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":63.2,"kv_cache_gb":0.15,"overhead_gb":3.03,"total_gb":66.38,"total_gb_q8_cache":66.33},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":63.2,"kv_cache_gb":0.25,"overhead_gb":3.03,"total_gb":66.48,"total_gb_q8_cache":66.38},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":63.2,"kv_cache_gb":0.45,"overhead_gb":3.03,"total_gb":66.68,"total_gb_q8_cache":66.48},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":63.2,"kv_cache_gb":0.86,"overhead_gb":3.03,"total_gb":67.08,"total_gb_q8_cache":66.68},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":262144,"weights_gb":63.2,"kv_cache_gb":1.66,"overhead_gb":3.03,"total_gb":67.89,"total_gb_q8_cache":67.08},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":33.5,"kv_cache_gb":0.08,"overhead_gb":1.84,"total_gb":35.41,"total_gb_q8_cache":35.4},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":33.5,"kv_cache_gb":0.1,"overhead_gb":1.84,"total_gb":35.44,"total_gb_q8_cache":35.41},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":33.5,"kv_cache_gb":0.15,"overhead_gb":1.84,"total_gb":35.49,"total_gb_q8_cache":35.44},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":33.5,"kv_cache_gb":0.25,"overhead_gb":1.84,"total_gb":35.59,"total_gb_q8_cache":35.49},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":33.5,"kv_cache_gb":0.45,"overhead_gb":1.84,"total_gb":35.79,"total_gb_q8_cache":35.59},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":33.5,"kv_cache_gb":0.86,"overhead_gb":1.84,"total_gb":36.19,"total_gb_q8_cache":35.79},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":262144,"weights_gb":33.5,"kv_cache_gb":1.66,"overhead_gb":1.84,"total_gb":37,"total_gb_q8_cache":36.19},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":25.91,"kv_cache_gb":0.08,"overhead_gb":1.54,"total_gb":27.52,"total_gb_q8_cache":27.51},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":25.91,"kv_cache_gb":0.1,"overhead_gb":1.54,"total_gb":27.55,"total_gb_q8_cache":27.52},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":25.91,"kv_cache_gb":0.15,"overhead_gb":1.54,"total_gb":27.6,"total_gb_q8_cache":27.55},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":25.91,"kv_cache_gb":0.25,"overhead_gb":1.54,"total_gb":27.7,"total_gb_q8_cache":27.6},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":25.91,"kv_cache_gb":0.45,"overhead_gb":1.54,"total_gb":27.9,"total_gb_q8_cache":27.7},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":25.91,"kv_cache_gb":0.86,"overhead_gb":1.54,"total_gb":28.3,"total_gb_q8_cache":27.9},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":262144,"weights_gb":25.91,"kv_cache_gb":1.66,"overhead_gb":1.54,"total_gb":29.11,"total_gb_q8_cache":28.3},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":22.44,"kv_cache_gb":0.08,"overhead_gb":1.4,"total_gb":23.91,"total_gb_q8_cache":23.9},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":22.44,"kv_cache_gb":0.1,"overhead_gb":1.4,"total_gb":23.93,"total_gb_q8_cache":23.91},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":22.44,"kv_cache_gb":0.15,"overhead_gb":1.4,"total_gb":23.98,"total_gb_q8_cache":23.93},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":22.44,"kv_cache_gb":0.25,"overhead_gb":1.4,"total_gb":24.08,"total_gb_q8_cache":23.98},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":22.44,"kv_cache_gb":0.45,"overhead_gb":1.4,"total_gb":24.29,"total_gb_q8_cache":24.08},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":22.44,"kv_cache_gb":0.86,"overhead_gb":1.4,"total_gb":24.69,"total_gb_q8_cache":24.29},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":262144,"weights_gb":22.44,"kv_cache_gb":1.66,"overhead_gb":1.4,"total_gb":25.49,"total_gb_q8_cache":24.69},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":18.33,"kv_cache_gb":0.08,"overhead_gb":1.23,"total_gb":19.64,"total_gb_q8_cache":19.62},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":18.33,"kv_cache_gb":0.1,"overhead_gb":1.23,"total_gb":19.66,"total_gb_q8_cache":19.64},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":18.33,"kv_cache_gb":0.15,"overhead_gb":1.23,"total_gb":19.71,"total_gb_q8_cache":19.66},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":18.33,"kv_cache_gb":0.25,"overhead_gb":1.23,"total_gb":19.81,"total_gb_q8_cache":19.71},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":18.33,"kv_cache_gb":0.45,"overhead_gb":1.23,"total_gb":20.01,"total_gb_q8_cache":19.81},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":18.33,"kv_cache_gb":0.86,"overhead_gb":1.23,"total_gb":20.42,"total_gb_q8_cache":20.01},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":262144,"weights_gb":18.33,"kv_cache_gb":1.66,"overhead_gb":1.23,"total_gb":21.22,"total_gb_q8_cache":20.42},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":14.85,"kv_cache_gb":0.08,"overhead_gb":1.09,"total_gb":16.02,"total_gb_q8_cache":16.01},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":14.85,"kv_cache_gb":0.1,"overhead_gb":1.09,"total_gb":16.05,"total_gb_q8_cache":16.02},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":14.85,"kv_cache_gb":0.15,"overhead_gb":1.09,"total_gb":16.1,"total_gb_q8_cache":16.05},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":14.85,"kv_cache_gb":0.25,"overhead_gb":1.09,"total_gb":16.2,"total_gb_q8_cache":16.1},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":14.85,"kv_cache_gb":0.45,"overhead_gb":1.09,"total_gb":16.4,"total_gb_q8_cache":16.2},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":14.85,"kv_cache_gb":0.86,"overhead_gb":1.09,"total_gb":16.8,"total_gb_q8_cache":16.4},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":262144,"weights_gb":14.85,"kv_cache_gb":1.66,"overhead_gb":1.09,"total_gb":17.61,"total_gb_q8_cache":16.8},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":65.6,"kv_cache_gb":1.07,"overhead_gb":3.12,"total_gb":69.8,"total_gb_q8_cache":69.26},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":65.6,"kv_cache_gb":2.15,"overhead_gb":3.12,"total_gb":70.87,"total_gb_q8_cache":69.8},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":65.6,"kv_cache_gb":4.29,"overhead_gb":3.12,"total_gb":73.02,"total_gb_q8_cache":70.87},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":65.6,"kv_cache_gb":8.59,"overhead_gb":3.12,"total_gb":77.31,"total_gb_q8_cache":73.02},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"FP16","bytes_per_param":2,"context_tokens":40960,"weights_gb":65.6,"kv_cache_gb":10.74,"overhead_gb":3.12,"total_gb":79.46,"total_gb_q8_cache":74.09},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":34.77,"kv_cache_gb":1.07,"overhead_gb":1.89,"total_gb":37.73,"total_gb_q8_cache":37.2},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":34.77,"kv_cache_gb":2.15,"overhead_gb":1.89,"total_gb":38.81,"total_gb_q8_cache":37.73},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":34.77,"kv_cache_gb":4.29,"overhead_gb":1.89,"total_gb":40.95,"total_gb_q8_cache":38.81},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":34.77,"kv_cache_gb":8.59,"overhead_gb":1.89,"total_gb":45.25,"total_gb_q8_cache":40.95},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":40960,"weights_gb":34.77,"kv_cache_gb":10.74,"overhead_gb":1.89,"total_gb":47.4,"total_gb_q8_cache":42.03},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":26.9,"kv_cache_gb":1.07,"overhead_gb":1.58,"total_gb":29.55,"total_gb_q8_cache":29.01},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":26.9,"kv_cache_gb":2.15,"overhead_gb":1.58,"total_gb":30.62,"total_gb_q8_cache":29.55},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":26.9,"kv_cache_gb":4.29,"overhead_gb":1.58,"total_gb":32.77,"total_gb_q8_cache":30.62},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":26.9,"kv_cache_gb":8.59,"overhead_gb":1.58,"total_gb":37.06,"total_gb_q8_cache":32.77},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":40960,"weights_gb":26.9,"kv_cache_gb":10.74,"overhead_gb":1.58,"total_gb":39.21,"total_gb_q8_cache":33.84},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":23.29,"kv_cache_gb":1.07,"overhead_gb":1.43,"total_gb":25.79,"total_gb_q8_cache":25.26},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":23.29,"kv_cache_gb":2.15,"overhead_gb":1.43,"total_gb":26.87,"total_gb_q8_cache":25.79},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":23.29,"kv_cache_gb":4.29,"overhead_gb":1.43,"total_gb":29.01,"total_gb_q8_cache":26.87},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":23.29,"kv_cache_gb":8.59,"overhead_gb":1.43,"total_gb":33.31,"total_gb_q8_cache":29.01},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":40960,"weights_gb":23.29,"kv_cache_gb":10.74,"overhead_gb":1.43,"total_gb":35.46,"total_gb_q8_cache":30.09},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":19.02,"kv_cache_gb":1.07,"overhead_gb":1.26,"total_gb":21.36,"total_gb_q8_cache":20.82},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":19.02,"kv_cache_gb":2.15,"overhead_gb":1.26,"total_gb":22.43,"total_gb_q8_cache":21.36},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":19.02,"kv_cache_gb":4.29,"overhead_gb":1.26,"total_gb":24.58,"total_gb_q8_cache":22.43},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":19.02,"kv_cache_gb":8.59,"overhead_gb":1.26,"total_gb":28.87,"total_gb_q8_cache":24.58},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":40960,"weights_gb":19.02,"kv_cache_gb":10.74,"overhead_gb":1.26,"total_gb":31.02,"total_gb_q8_cache":25.65},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":15.42,"kv_cache_gb":1.07,"overhead_gb":1.12,"total_gb":17.61,"total_gb_q8_cache":17.07},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":15.42,"kv_cache_gb":2.15,"overhead_gb":1.12,"total_gb":18.68,"total_gb_q8_cache":17.61},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":15.42,"kv_cache_gb":4.29,"overhead_gb":1.12,"total_gb":20.83,"total_gb_q8_cache":18.68},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":15.42,"kv_cache_gb":8.59,"overhead_gb":1.12,"total_gb":25.12,"total_gb_q8_cache":20.83},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":40960,"weights_gb":15.42,"kv_cache_gb":10.74,"overhead_gb":1.12,"total_gb":27.27,"total_gb_q8_cache":21.9},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":65.6,"kv_cache_gb":1.07,"overhead_gb":3.12,"total_gb":69.8,"total_gb_q8_cache":69.26},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":65.6,"kv_cache_gb":2.15,"overhead_gb":3.12,"total_gb":70.87,"total_gb_q8_cache":69.8},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":65.6,"kv_cache_gb":4.29,"overhead_gb":3.12,"total_gb":73.02,"total_gb_q8_cache":70.87},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":65.6,"kv_cache_gb":8.59,"overhead_gb":3.12,"total_gb":77.31,"total_gb_q8_cache":73.02},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":34.77,"kv_cache_gb":1.07,"overhead_gb":1.89,"total_gb":37.73,"total_gb_q8_cache":37.2},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":34.77,"kv_cache_gb":2.15,"overhead_gb":1.89,"total_gb":38.81,"total_gb_q8_cache":37.73},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":34.77,"kv_cache_gb":4.29,"overhead_gb":1.89,"total_gb":40.95,"total_gb_q8_cache":38.81},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":34.77,"kv_cache_gb":8.59,"overhead_gb":1.89,"total_gb":45.25,"total_gb_q8_cache":40.95},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":26.9,"kv_cache_gb":1.07,"overhead_gb":1.58,"total_gb":29.55,"total_gb_q8_cache":29.01},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":26.9,"kv_cache_gb":2.15,"overhead_gb":1.58,"total_gb":30.62,"total_gb_q8_cache":29.55},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":26.9,"kv_cache_gb":4.29,"overhead_gb":1.58,"total_gb":32.77,"total_gb_q8_cache":30.62},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":26.9,"kv_cache_gb":8.59,"overhead_gb":1.58,"total_gb":37.06,"total_gb_q8_cache":32.77},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":23.29,"kv_cache_gb":1.07,"overhead_gb":1.43,"total_gb":25.79,"total_gb_q8_cache":25.26},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":23.29,"kv_cache_gb":2.15,"overhead_gb":1.43,"total_gb":26.87,"total_gb_q8_cache":25.79},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":23.29,"kv_cache_gb":4.29,"overhead_gb":1.43,"total_gb":29.01,"total_gb_q8_cache":26.87},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":23.29,"kv_cache_gb":8.59,"overhead_gb":1.43,"total_gb":33.31,"total_gb_q8_cache":29.01},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":19.02,"kv_cache_gb":1.07,"overhead_gb":1.26,"total_gb":21.36,"total_gb_q8_cache":20.82},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":19.02,"kv_cache_gb":2.15,"overhead_gb":1.26,"total_gb":22.43,"total_gb_q8_cache":21.36},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":19.02,"kv_cache_gb":4.29,"overhead_gb":1.26,"total_gb":24.58,"total_gb_q8_cache":22.43},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":19.02,"kv_cache_gb":8.59,"overhead_gb":1.26,"total_gb":28.87,"total_gb_q8_cache":24.58},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":15.42,"kv_cache_gb":1.07,"overhead_gb":1.12,"total_gb":17.61,"total_gb_q8_cache":17.07},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":15.42,"kv_cache_gb":2.15,"overhead_gb":1.12,"total_gb":18.68,"total_gb_q8_cache":17.61},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":15.42,"kv_cache_gb":4.29,"overhead_gb":1.12,"total_gb":20.83,"total_gb_q8_cache":18.68},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":15.42,"kv_cache_gb":8.59,"overhead_gb":1.12,"total_gb":25.12,"total_gb_q8_cache":20.83},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":65.6,"kv_cache_gb":1.07,"overhead_gb":3.12,"total_gb":69.8,"total_gb_q8_cache":69.26},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":65.6,"kv_cache_gb":2.15,"overhead_gb":3.12,"total_gb":70.87,"total_gb_q8_cache":69.8},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":65.6,"kv_cache_gb":4.29,"overhead_gb":3.12,"total_gb":73.02,"total_gb_q8_cache":70.87},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":65.6,"kv_cache_gb":8.59,"overhead_gb":3.12,"total_gb":77.31,"total_gb_q8_cache":73.02},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":65.6,"kv_cache_gb":17.18,"overhead_gb":3.12,"total_gb":85.9,"total_gb_q8_cache":77.31},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":65.6,"kv_cache_gb":34.36,"overhead_gb":3.12,"total_gb":103.08,"total_gb_q8_cache":85.9},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":34.77,"kv_cache_gb":1.07,"overhead_gb":1.89,"total_gb":37.73,"total_gb_q8_cache":37.2},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":34.77,"kv_cache_gb":2.15,"overhead_gb":1.89,"total_gb":38.81,"total_gb_q8_cache":37.73},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":34.77,"kv_cache_gb":4.29,"overhead_gb":1.89,"total_gb":40.95,"total_gb_q8_cache":38.81},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":34.77,"kv_cache_gb":8.59,"overhead_gb":1.89,"total_gb":45.25,"total_gb_q8_cache":40.95},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":34.77,"kv_cache_gb":17.18,"overhead_gb":1.89,"total_gb":53.84,"total_gb_q8_cache":45.25},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":34.77,"kv_cache_gb":34.36,"overhead_gb":1.89,"total_gb":71.02,"total_gb_q8_cache":53.84},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":26.9,"kv_cache_gb":1.07,"overhead_gb":1.58,"total_gb":29.55,"total_gb_q8_cache":29.01},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":26.9,"kv_cache_gb":2.15,"overhead_gb":1.58,"total_gb":30.62,"total_gb_q8_cache":29.55},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":26.9,"kv_cache_gb":4.29,"overhead_gb":1.58,"total_gb":32.77,"total_gb_q8_cache":30.62},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":26.9,"kv_cache_gb":8.59,"overhead_gb":1.58,"total_gb":37.06,"total_gb_q8_cache":32.77},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":26.9,"kv_cache_gb":17.18,"overhead_gb":1.58,"total_gb":45.65,"total_gb_q8_cache":37.06},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":26.9,"kv_cache_gb":34.36,"overhead_gb":1.58,"total_gb":62.83,"total_gb_q8_cache":45.65},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":23.29,"kv_cache_gb":1.07,"overhead_gb":1.43,"total_gb":25.79,"total_gb_q8_cache":25.26},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":23.29,"kv_cache_gb":2.15,"overhead_gb":1.43,"total_gb":26.87,"total_gb_q8_cache":25.79},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":23.29,"kv_cache_gb":4.29,"overhead_gb":1.43,"total_gb":29.01,"total_gb_q8_cache":26.87},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":23.29,"kv_cache_gb":8.59,"overhead_gb":1.43,"total_gb":33.31,"total_gb_q8_cache":29.01},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":23.29,"kv_cache_gb":17.18,"overhead_gb":1.43,"total_gb":41.9,"total_gb_q8_cache":33.31},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":23.29,"kv_cache_gb":34.36,"overhead_gb":1.43,"total_gb":59.08,"total_gb_q8_cache":41.9},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":19.02,"kv_cache_gb":1.07,"overhead_gb":1.26,"total_gb":21.36,"total_gb_q8_cache":20.82},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":19.02,"kv_cache_gb":2.15,"overhead_gb":1.26,"total_gb":22.43,"total_gb_q8_cache":21.36},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":19.02,"kv_cache_gb":4.29,"overhead_gb":1.26,"total_gb":24.58,"total_gb_q8_cache":22.43},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":19.02,"kv_cache_gb":8.59,"overhead_gb":1.26,"total_gb":28.87,"total_gb_q8_cache":24.58},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":19.02,"kv_cache_gb":17.18,"overhead_gb":1.26,"total_gb":37.46,"total_gb_q8_cache":28.87},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":19.02,"kv_cache_gb":34.36,"overhead_gb":1.26,"total_gb":54.64,"total_gb_q8_cache":37.46},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":15.42,"kv_cache_gb":1.07,"overhead_gb":1.12,"total_gb":17.61,"total_gb_q8_cache":17.07},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":15.42,"kv_cache_gb":2.15,"overhead_gb":1.12,"total_gb":18.68,"total_gb_q8_cache":17.61},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":15.42,"kv_cache_gb":4.29,"overhead_gb":1.12,"total_gb":20.83,"total_gb_q8_cache":18.68},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":15.42,"kv_cache_gb":8.59,"overhead_gb":1.12,"total_gb":25.12,"total_gb_q8_cache":20.83},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":15.42,"kv_cache_gb":17.18,"overhead_gb":1.12,"total_gb":33.71,"total_gb_q8_cache":25.12},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":15.42,"kv_cache_gb":34.36,"overhead_gb":1.12,"total_gb":50.89,"total_gb_q8_cache":33.71},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":72,"kv_cache_gb":0.14,"overhead_gb":3.38,"total_gb":75.52,"total_gb_q8_cache":75.48},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":72,"kv_cache_gb":0.23,"overhead_gb":3.38,"total_gb":75.61,"total_gb_q8_cache":75.52},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":72,"kv_cache_gb":0.4,"overhead_gb":3.38,"total_gb":75.78,"total_gb_q8_cache":75.61},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":72,"kv_cache_gb":0.73,"overhead_gb":3.38,"total_gb":76.11,"total_gb_q8_cache":75.78},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":72,"kv_cache_gb":1.4,"overhead_gb":3.38,"total_gb":76.78,"total_gb_q8_cache":76.11},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":72,"kv_cache_gb":2.74,"overhead_gb":3.38,"total_gb":78.12,"total_gb_q8_cache":76.78},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":262144,"weights_gb":72,"kv_cache_gb":5.43,"overhead_gb":3.38,"total_gb":80.81,"total_gb_q8_cache":78.12},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":38.16,"kv_cache_gb":0.14,"overhead_gb":2.03,"total_gb":40.33,"total_gb_q8_cache":40.29},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":38.16,"kv_cache_gb":0.23,"overhead_gb":2.03,"total_gb":40.41,"total_gb_q8_cache":40.33},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":38.16,"kv_cache_gb":0.4,"overhead_gb":2.03,"total_gb":40.58,"total_gb_q8_cache":40.41},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":38.16,"kv_cache_gb":0.73,"overhead_gb":2.03,"total_gb":40.92,"total_gb_q8_cache":40.58},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":38.16,"kv_cache_gb":1.4,"overhead_gb":2.03,"total_gb":41.59,"total_gb_q8_cache":40.92},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":38.16,"kv_cache_gb":2.74,"overhead_gb":2.03,"total_gb":42.93,"total_gb_q8_cache":41.59},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":262144,"weights_gb":38.16,"kv_cache_gb":5.43,"overhead_gb":2.03,"total_gb":45.62,"total_gb_q8_cache":42.93},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":29.52,"kv_cache_gb":0.14,"overhead_gb":1.68,"total_gb":31.34,"total_gb_q8_cache":31.3},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":29.52,"kv_cache_gb":0.23,"overhead_gb":1.68,"total_gb":31.43,"total_gb_q8_cache":31.34},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":29.52,"kv_cache_gb":0.4,"overhead_gb":1.68,"total_gb":31.6,"total_gb_q8_cache":31.43},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":29.52,"kv_cache_gb":0.73,"overhead_gb":1.68,"total_gb":31.93,"total_gb_q8_cache":31.6},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":29.52,"kv_cache_gb":1.4,"overhead_gb":1.68,"total_gb":32.6,"total_gb_q8_cache":31.93},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":29.52,"kv_cache_gb":2.74,"overhead_gb":1.68,"total_gb":33.95,"total_gb_q8_cache":32.6},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":262144,"weights_gb":29.52,"kv_cache_gb":5.43,"overhead_gb":1.68,"total_gb":36.63,"total_gb_q8_cache":33.95},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":25.56,"kv_cache_gb":0.14,"overhead_gb":1.52,"total_gb":27.23,"total_gb_q8_cache":27.18},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":25.56,"kv_cache_gb":0.23,"overhead_gb":1.52,"total_gb":27.31,"total_gb_q8_cache":27.23},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":25.56,"kv_cache_gb":0.4,"overhead_gb":1.52,"total_gb":27.48,"total_gb_q8_cache":27.31},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":25.56,"kv_cache_gb":0.73,"overhead_gb":1.52,"total_gb":27.81,"total_gb_q8_cache":27.48},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":25.56,"kv_cache_gb":1.4,"overhead_gb":1.52,"total_gb":28.48,"total_gb_q8_cache":27.81},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":25.56,"kv_cache_gb":2.74,"overhead_gb":1.52,"total_gb":29.83,"total_gb_q8_cache":28.48},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":262144,"weights_gb":25.56,"kv_cache_gb":5.43,"overhead_gb":1.52,"total_gb":32.51,"total_gb_q8_cache":29.83},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":20.88,"kv_cache_gb":0.14,"overhead_gb":1.34,"total_gb":22.36,"total_gb_q8_cache":22.32},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":20.88,"kv_cache_gb":0.23,"overhead_gb":1.34,"total_gb":22.44,"total_gb_q8_cache":22.36},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":20.88,"kv_cache_gb":0.4,"overhead_gb":1.34,"total_gb":22.61,"total_gb_q8_cache":22.44},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":20.88,"kv_cache_gb":0.73,"overhead_gb":1.34,"total_gb":22.95,"total_gb_q8_cache":22.61},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":20.88,"kv_cache_gb":1.4,"overhead_gb":1.34,"total_gb":23.62,"total_gb_q8_cache":22.95},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":20.88,"kv_cache_gb":2.74,"overhead_gb":1.34,"total_gb":24.96,"total_gb_q8_cache":23.62},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":262144,"weights_gb":20.88,"kv_cache_gb":5.43,"overhead_gb":1.34,"total_gb":27.64,"total_gb_q8_cache":24.96},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":16.92,"kv_cache_gb":0.14,"overhead_gb":1.18,"total_gb":18.24,"total_gb_q8_cache":18.2},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":16.92,"kv_cache_gb":0.23,"overhead_gb":1.18,"total_gb":18.32,"total_gb_q8_cache":18.24},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":16.92,"kv_cache_gb":0.4,"overhead_gb":1.18,"total_gb":18.49,"total_gb_q8_cache":18.32},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":16.92,"kv_cache_gb":0.73,"overhead_gb":1.18,"total_gb":18.83,"total_gb_q8_cache":18.49},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":16.92,"kv_cache_gb":1.4,"overhead_gb":1.18,"total_gb":19.5,"total_gb_q8_cache":18.83},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":16.92,"kv_cache_gb":2.74,"overhead_gb":1.18,"total_gb":20.84,"total_gb_q8_cache":19.5},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":262144,"weights_gb":16.92,"kv_cache_gb":5.43,"overhead_gb":1.18,"total_gb":23.53,"total_gb_q8_cache":20.84},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":141.2,"kv_cache_gb":1.34,"overhead_gb":6.15,"total_gb":148.69,"total_gb_q8_cache":148.02},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":141.2,"kv_cache_gb":2.68,"overhead_gb":6.15,"total_gb":150.03,"total_gb_q8_cache":148.69},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":141.2,"kv_cache_gb":5.37,"overhead_gb":6.15,"total_gb":152.72,"total_gb_q8_cache":150.03},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":141.2,"kv_cache_gb":10.74,"overhead_gb":6.15,"total_gb":158.09,"total_gb_q8_cache":152.72},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":141.2,"kv_cache_gb":21.47,"overhead_gb":6.15,"total_gb":168.82,"total_gb_q8_cache":158.09},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":141.2,"kv_cache_gb":42.95,"overhead_gb":6.15,"total_gb":190.3,"total_gb_q8_cache":168.82},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":74.84,"kv_cache_gb":1.34,"overhead_gb":3.49,"total_gb":79.67,"total_gb_q8_cache":79},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":74.84,"kv_cache_gb":2.68,"overhead_gb":3.49,"total_gb":81.01,"total_gb_q8_cache":79.67},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":74.84,"kv_cache_gb":5.37,"overhead_gb":3.49,"total_gb":83.7,"total_gb_q8_cache":81.01},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":74.84,"kv_cache_gb":10.74,"overhead_gb":3.49,"total_gb":89.07,"total_gb_q8_cache":83.7},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":74.84,"kv_cache_gb":21.47,"overhead_gb":3.49,"total_gb":99.8,"total_gb_q8_cache":89.07},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":74.84,"kv_cache_gb":42.95,"overhead_gb":3.49,"total_gb":121.28,"total_gb_q8_cache":99.8},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":57.89,"kv_cache_gb":1.34,"overhead_gb":2.82,"total_gb":62.05,"total_gb_q8_cache":61.38},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":57.89,"kv_cache_gb":2.68,"overhead_gb":2.82,"total_gb":63.39,"total_gb_q8_cache":62.05},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":57.89,"kv_cache_gb":5.37,"overhead_gb":2.82,"total_gb":66.08,"total_gb_q8_cache":63.39},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":57.89,"kv_cache_gb":10.74,"overhead_gb":2.82,"total_gb":71.45,"total_gb_q8_cache":66.08},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":57.89,"kv_cache_gb":21.47,"overhead_gb":2.82,"total_gb":82.18,"total_gb_q8_cache":71.45},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":57.89,"kv_cache_gb":42.95,"overhead_gb":2.82,"total_gb":103.66,"total_gb_q8_cache":82.18},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":50.13,"kv_cache_gb":1.34,"overhead_gb":2.51,"total_gb":53.97,"total_gb_q8_cache":53.3},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":50.13,"kv_cache_gb":2.68,"overhead_gb":2.51,"total_gb":55.32,"total_gb_q8_cache":53.97},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":50.13,"kv_cache_gb":5.37,"overhead_gb":2.51,"total_gb":58,"total_gb_q8_cache":55.32},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":50.13,"kv_cache_gb":10.74,"overhead_gb":2.51,"total_gb":63.37,"total_gb_q8_cache":58},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":50.13,"kv_cache_gb":21.47,"overhead_gb":2.51,"total_gb":74.11,"total_gb_q8_cache":63.37},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":50.13,"kv_cache_gb":42.95,"overhead_gb":2.51,"total_gb":95.58,"total_gb_q8_cache":74.11},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":40.95,"kv_cache_gb":1.34,"overhead_gb":2.14,"total_gb":44.43,"total_gb_q8_cache":43.76},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":40.95,"kv_cache_gb":2.68,"overhead_gb":2.14,"total_gb":45.77,"total_gb_q8_cache":44.43},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":40.95,"kv_cache_gb":5.37,"overhead_gb":2.14,"total_gb":48.45,"total_gb_q8_cache":45.77},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":40.95,"kv_cache_gb":10.74,"overhead_gb":2.14,"total_gb":53.82,"total_gb_q8_cache":48.45},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":40.95,"kv_cache_gb":21.47,"overhead_gb":2.14,"total_gb":64.56,"total_gb_q8_cache":53.82},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":40.95,"kv_cache_gb":42.95,"overhead_gb":2.14,"total_gb":86.04,"total_gb_q8_cache":64.56},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":33.18,"kv_cache_gb":1.34,"overhead_gb":1.83,"total_gb":36.35,"total_gb_q8_cache":35.68},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":33.18,"kv_cache_gb":2.68,"overhead_gb":1.83,"total_gb":37.69,"total_gb_q8_cache":36.35},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":33.18,"kv_cache_gb":5.37,"overhead_gb":1.83,"total_gb":40.38,"total_gb_q8_cache":37.69},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":33.18,"kv_cache_gb":10.74,"overhead_gb":1.83,"total_gb":45.75,"total_gb_q8_cache":40.38},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":33.18,"kv_cache_gb":21.47,"overhead_gb":1.83,"total_gb":56.48,"total_gb_q8_cache":45.75},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":33.18,"kv_cache_gb":42.95,"overhead_gb":1.83,"total_gb":77.96,"total_gb_q8_cache":56.48},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":141.2,"kv_cache_gb":1.34,"overhead_gb":6.15,"total_gb":148.69,"total_gb_q8_cache":148.02},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":141.2,"kv_cache_gb":2.68,"overhead_gb":6.15,"total_gb":150.03,"total_gb_q8_cache":148.69},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":141.2,"kv_cache_gb":5.37,"overhead_gb":6.15,"total_gb":152.72,"total_gb_q8_cache":150.03},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":141.2,"kv_cache_gb":10.74,"overhead_gb":6.15,"total_gb":158.09,"total_gb_q8_cache":152.72},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":141.2,"kv_cache_gb":21.47,"overhead_gb":6.15,"total_gb":168.82,"total_gb_q8_cache":158.09},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":141.2,"kv_cache_gb":42.95,"overhead_gb":6.15,"total_gb":190.3,"total_gb_q8_cache":168.82},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":74.84,"kv_cache_gb":1.34,"overhead_gb":3.49,"total_gb":79.67,"total_gb_q8_cache":79},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":74.84,"kv_cache_gb":2.68,"overhead_gb":3.49,"total_gb":81.01,"total_gb_q8_cache":79.67},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":74.84,"kv_cache_gb":5.37,"overhead_gb":3.49,"total_gb":83.7,"total_gb_q8_cache":81.01},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":74.84,"kv_cache_gb":10.74,"overhead_gb":3.49,"total_gb":89.07,"total_gb_q8_cache":83.7},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":74.84,"kv_cache_gb":21.47,"overhead_gb":3.49,"total_gb":99.8,"total_gb_q8_cache":89.07},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":74.84,"kv_cache_gb":42.95,"overhead_gb":3.49,"total_gb":121.28,"total_gb_q8_cache":99.8},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":57.89,"kv_cache_gb":1.34,"overhead_gb":2.82,"total_gb":62.05,"total_gb_q8_cache":61.38},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":57.89,"kv_cache_gb":2.68,"overhead_gb":2.82,"total_gb":63.39,"total_gb_q8_cache":62.05},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":57.89,"kv_cache_gb":5.37,"overhead_gb":2.82,"total_gb":66.08,"total_gb_q8_cache":63.39},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":57.89,"kv_cache_gb":10.74,"overhead_gb":2.82,"total_gb":71.45,"total_gb_q8_cache":66.08},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":57.89,"kv_cache_gb":21.47,"overhead_gb":2.82,"total_gb":82.18,"total_gb_q8_cache":71.45},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":57.89,"kv_cache_gb":42.95,"overhead_gb":2.82,"total_gb":103.66,"total_gb_q8_cache":82.18},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":50.13,"kv_cache_gb":1.34,"overhead_gb":2.51,"total_gb":53.97,"total_gb_q8_cache":53.3},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":50.13,"kv_cache_gb":2.68,"overhead_gb":2.51,"total_gb":55.32,"total_gb_q8_cache":53.97},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":50.13,"kv_cache_gb":5.37,"overhead_gb":2.51,"total_gb":58,"total_gb_q8_cache":55.32},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":50.13,"kv_cache_gb":10.74,"overhead_gb":2.51,"total_gb":63.37,"total_gb_q8_cache":58},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":50.13,"kv_cache_gb":21.47,"overhead_gb":2.51,"total_gb":74.11,"total_gb_q8_cache":63.37},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":50.13,"kv_cache_gb":42.95,"overhead_gb":2.51,"total_gb":95.58,"total_gb_q8_cache":74.11},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":40.95,"kv_cache_gb":1.34,"overhead_gb":2.14,"total_gb":44.43,"total_gb_q8_cache":43.76},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":40.95,"kv_cache_gb":2.68,"overhead_gb":2.14,"total_gb":45.77,"total_gb_q8_cache":44.43},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":40.95,"kv_cache_gb":5.37,"overhead_gb":2.14,"total_gb":48.45,"total_gb_q8_cache":45.77},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":40.95,"kv_cache_gb":10.74,"overhead_gb":2.14,"total_gb":53.82,"total_gb_q8_cache":48.45},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":40.95,"kv_cache_gb":21.47,"overhead_gb":2.14,"total_gb":64.56,"total_gb_q8_cache":53.82},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":40.95,"kv_cache_gb":42.95,"overhead_gb":2.14,"total_gb":86.04,"total_gb_q8_cache":64.56},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":33.18,"kv_cache_gb":1.34,"overhead_gb":1.83,"total_gb":36.35,"total_gb_q8_cache":35.68},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":33.18,"kv_cache_gb":2.68,"overhead_gb":1.83,"total_gb":37.69,"total_gb_q8_cache":36.35},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":33.18,"kv_cache_gb":5.37,"overhead_gb":1.83,"total_gb":40.38,"total_gb_q8_cache":37.69},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":33.18,"kv_cache_gb":10.74,"overhead_gb":1.83,"total_gb":45.75,"total_gb_q8_cache":40.38},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":33.18,"kv_cache_gb":21.47,"overhead_gb":1.83,"total_gb":56.48,"total_gb_q8_cache":45.75},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":33.18,"kv_cache_gb":42.95,"overhead_gb":1.83,"total_gb":77.96,"total_gb_q8_cache":56.48},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":159.4,"kv_cache_gb":0.18,"overhead_gb":6.88,"total_gb":166.46,"total_gb_q8_cache":166.41},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":159.4,"kv_cache_gb":0.28,"overhead_gb":6.88,"total_gb":166.56,"total_gb_q8_cache":166.46},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":159.4,"kv_cache_gb":0.48,"overhead_gb":6.88,"total_gb":166.76,"total_gb_q8_cache":166.56},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":159.4,"kv_cache_gb":0.89,"overhead_gb":6.88,"total_gb":167.16,"total_gb_q8_cache":166.76},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":159.4,"kv_cache_gb":1.69,"overhead_gb":6.88,"total_gb":167.97,"total_gb_q8_cache":167.16},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":159.4,"kv_cache_gb":3.3,"overhead_gb":6.88,"total_gb":169.58,"total_gb_q8_cache":167.97},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":262144,"weights_gb":159.4,"kv_cache_gb":6.52,"overhead_gb":6.88,"total_gb":172.8,"total_gb_q8_cache":169.58},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":84.48,"kv_cache_gb":0.18,"overhead_gb":3.88,"total_gb":88.54,"total_gb_q8_cache":88.49},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":84.48,"kv_cache_gb":0.28,"overhead_gb":3.88,"total_gb":88.64,"total_gb_q8_cache":88.54},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":84.48,"kv_cache_gb":0.48,"overhead_gb":3.88,"total_gb":88.84,"total_gb_q8_cache":88.64},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":84.48,"kv_cache_gb":0.89,"overhead_gb":3.88,"total_gb":89.25,"total_gb_q8_cache":88.84},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":84.48,"kv_cache_gb":1.69,"overhead_gb":3.88,"total_gb":90.05,"total_gb_q8_cache":89.25},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":84.48,"kv_cache_gb":3.3,"overhead_gb":3.88,"total_gb":91.66,"total_gb_q8_cache":90.05},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":262144,"weights_gb":84.48,"kv_cache_gb":6.52,"overhead_gb":3.88,"total_gb":94.88,"total_gb_q8_cache":91.66},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":65.35,"kv_cache_gb":0.18,"overhead_gb":3.11,"total_gb":68.65,"total_gb_q8_cache":68.6},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":65.35,"kv_cache_gb":0.28,"overhead_gb":3.11,"total_gb":68.75,"total_gb_q8_cache":68.65},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":65.35,"kv_cache_gb":0.48,"overhead_gb":3.11,"total_gb":68.95,"total_gb_q8_cache":68.75},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":65.35,"kv_cache_gb":0.89,"overhead_gb":3.11,"total_gb":69.35,"total_gb_q8_cache":68.95},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":65.35,"kv_cache_gb":1.69,"overhead_gb":3.11,"total_gb":70.16,"total_gb_q8_cache":69.35},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":65.35,"kv_cache_gb":3.3,"overhead_gb":3.11,"total_gb":71.77,"total_gb_q8_cache":70.16},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":262144,"weights_gb":65.35,"kv_cache_gb":6.52,"overhead_gb":3.11,"total_gb":74.99,"total_gb_q8_cache":71.77},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":56.59,"kv_cache_gb":0.18,"overhead_gb":2.76,"total_gb":59.53,"total_gb_q8_cache":59.48},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":56.59,"kv_cache_gb":0.28,"overhead_gb":2.76,"total_gb":59.63,"total_gb_q8_cache":59.53},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":56.59,"kv_cache_gb":0.48,"overhead_gb":2.76,"total_gb":59.83,"total_gb_q8_cache":59.63},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":56.59,"kv_cache_gb":0.89,"overhead_gb":2.76,"total_gb":60.24,"total_gb_q8_cache":59.83},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":56.59,"kv_cache_gb":1.69,"overhead_gb":2.76,"total_gb":61.04,"total_gb_q8_cache":60.24},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":56.59,"kv_cache_gb":3.3,"overhead_gb":2.76,"total_gb":62.65,"total_gb_q8_cache":61.04},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":262144,"weights_gb":56.59,"kv_cache_gb":6.52,"overhead_gb":2.76,"total_gb":65.87,"total_gb_q8_cache":62.65},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":46.23,"kv_cache_gb":0.18,"overhead_gb":2.35,"total_gb":48.76,"total_gb_q8_cache":48.71},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":46.23,"kv_cache_gb":0.28,"overhead_gb":2.35,"total_gb":48.86,"total_gb_q8_cache":48.76},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":46.23,"kv_cache_gb":0.48,"overhead_gb":2.35,"total_gb":49.06,"total_gb_q8_cache":48.86},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":46.23,"kv_cache_gb":0.89,"overhead_gb":2.35,"total_gb":49.46,"total_gb_q8_cache":49.06},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":46.23,"kv_cache_gb":1.69,"overhead_gb":2.35,"total_gb":50.27,"total_gb_q8_cache":49.46},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":46.23,"kv_cache_gb":3.3,"overhead_gb":2.35,"total_gb":51.88,"total_gb_q8_cache":50.27},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":262144,"weights_gb":46.23,"kv_cache_gb":6.52,"overhead_gb":2.35,"total_gb":55.1,"total_gb_q8_cache":51.88},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":37.46,"kv_cache_gb":0.18,"overhead_gb":2,"total_gb":39.64,"total_gb_q8_cache":39.59},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":37.46,"kv_cache_gb":0.28,"overhead_gb":2,"total_gb":39.74,"total_gb_q8_cache":39.64},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":37.46,"kv_cache_gb":0.48,"overhead_gb":2,"total_gb":39.94,"total_gb_q8_cache":39.74},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":37.46,"kv_cache_gb":0.89,"overhead_gb":2,"total_gb":40.34,"total_gb_q8_cache":39.94},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":37.46,"kv_cache_gb":1.69,"overhead_gb":2,"total_gb":41.15,"total_gb_q8_cache":40.34},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":37.46,"kv_cache_gb":3.3,"overhead_gb":2,"total_gb":42.76,"total_gb_q8_cache":41.15},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":262144,"weights_gb":37.46,"kv_cache_gb":6.52,"overhead_gb":2,"total_gb":45.98,"total_gb_q8_cache":42.76},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":234,"kv_cache_gb":0.16,"overhead_gb":9.86,"total_gb":244.02,"total_gb_q8_cache":243.94},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":234,"kv_cache_gb":0.31,"overhead_gb":9.86,"total_gb":244.17,"total_gb_q8_cache":244.01},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":234,"kv_cache_gb":0.61,"overhead_gb":9.86,"total_gb":244.47,"total_gb_q8_cache":244.16},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":234,"kv_cache_gb":1.21,"overhead_gb":9.86,"total_gb":245.07,"total_gb_q8_cache":244.47},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":234,"kv_cache_gb":2.42,"overhead_gb":9.86,"total_gb":246.28,"total_gb_q8_cache":245.07},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":234,"kv_cache_gb":4.84,"overhead_gb":9.86,"total_gb":248.7,"total_gb_q8_cache":246.28},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":124.02,"kv_cache_gb":0.16,"overhead_gb":5.46,"total_gb":129.64,"total_gb_q8_cache":129.56},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":124.02,"kv_cache_gb":0.31,"overhead_gb":5.46,"total_gb":129.79,"total_gb_q8_cache":129.63},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":124.02,"kv_cache_gb":0.61,"overhead_gb":5.46,"total_gb":130.09,"total_gb_q8_cache":129.79},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":124.02,"kv_cache_gb":1.21,"overhead_gb":5.46,"total_gb":130.69,"total_gb_q8_cache":130.09},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":124.02,"kv_cache_gb":2.42,"overhead_gb":5.46,"total_gb":131.9,"total_gb_q8_cache":130.69},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":124.02,"kv_cache_gb":4.84,"overhead_gb":5.46,"total_gb":134.32,"total_gb_q8_cache":131.9},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":95.94,"kv_cache_gb":0.16,"overhead_gb":4.34,"total_gb":100.43,"total_gb_q8_cache":100.36},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":95.94,"kv_cache_gb":0.31,"overhead_gb":4.34,"total_gb":100.58,"total_gb_q8_cache":100.43},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":95.94,"kv_cache_gb":0.61,"overhead_gb":4.34,"total_gb":100.89,"total_gb_q8_cache":100.58},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":95.94,"kv_cache_gb":1.21,"overhead_gb":4.34,"total_gb":101.49,"total_gb_q8_cache":100.88},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":95.94,"kv_cache_gb":2.42,"overhead_gb":4.34,"total_gb":102.7,"total_gb_q8_cache":101.49},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":95.94,"kv_cache_gb":4.84,"overhead_gb":4.34,"total_gb":105.11,"total_gb_q8_cache":102.7},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":83.07,"kv_cache_gb":0.16,"overhead_gb":3.82,"total_gb":87.05,"total_gb_q8_cache":86.97},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":83.07,"kv_cache_gb":0.31,"overhead_gb":3.82,"total_gb":87.2,"total_gb_q8_cache":87.05},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":83.07,"kv_cache_gb":0.61,"overhead_gb":3.82,"total_gb":87.5,"total_gb_q8_cache":87.2},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":83.07,"kv_cache_gb":1.21,"overhead_gb":3.82,"total_gb":88.11,"total_gb_q8_cache":87.5},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":83.07,"kv_cache_gb":2.42,"overhead_gb":3.82,"total_gb":89.31,"total_gb_q8_cache":88.1},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":83.07,"kv_cache_gb":4.84,"overhead_gb":3.82,"total_gb":91.73,"total_gb_q8_cache":89.31},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":67.86,"kv_cache_gb":0.16,"overhead_gb":3.21,"total_gb":71.23,"total_gb_q8_cache":71.15},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":67.86,"kv_cache_gb":0.31,"overhead_gb":3.21,"total_gb":71.38,"total_gb_q8_cache":71.23},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":67.86,"kv_cache_gb":0.61,"overhead_gb":3.21,"total_gb":71.68,"total_gb_q8_cache":71.38},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":67.86,"kv_cache_gb":1.21,"overhead_gb":3.21,"total_gb":72.29,"total_gb_q8_cache":71.68},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":67.86,"kv_cache_gb":2.42,"overhead_gb":3.21,"total_gb":73.5,"total_gb_q8_cache":72.28},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":67.86,"kv_cache_gb":4.84,"overhead_gb":3.21,"total_gb":75.91,"total_gb_q8_cache":73.49},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":54.99,"kv_cache_gb":0.16,"overhead_gb":2.7,"total_gb":57.85,"total_gb_q8_cache":57.77},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":54.99,"kv_cache_gb":0.31,"overhead_gb":2.7,"total_gb":58,"total_gb_q8_cache":57.84},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":54.99,"kv_cache_gb":0.61,"overhead_gb":2.7,"total_gb":58.3,"total_gb_q8_cache":57.99},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":54.99,"kv_cache_gb":1.21,"overhead_gb":2.7,"total_gb":58.9,"total_gb_q8_cache":58.3},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":54.99,"kv_cache_gb":2.42,"overhead_gb":2.7,"total_gb":60.11,"total_gb_q8_cache":58.9},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":54.99,"kv_cache_gb":4.84,"overhead_gb":2.7,"total_gb":62.53,"total_gb_q8_cache":60.11},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":4096,"weights_gb":238,"kv_cache_gb":0.09,"overhead_gb":10.02,"total_gb":248.11,"total_gb_q8_cache":248.07},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":8192,"weights_gb":238,"kv_cache_gb":0.19,"overhead_gb":10.02,"total_gb":248.21,"total_gb_q8_cache":248.11},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":16384,"weights_gb":238,"kv_cache_gb":0.38,"overhead_gb":10.02,"total_gb":248.4,"total_gb_q8_cache":248.21},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":32768,"weights_gb":238,"kv_cache_gb":0.75,"overhead_gb":10.02,"total_gb":248.77,"total_gb_q8_cache":248.4},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":65536,"weights_gb":238,"kv_cache_gb":1.51,"overhead_gb":10.02,"total_gb":249.53,"total_gb_q8_cache":248.77},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":131072,"weights_gb":238,"kv_cache_gb":3.02,"overhead_gb":10.02,"total_gb":251.04,"total_gb_q8_cache":249.53},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"FP16","bytes_per_param":2,"context_tokens":262144,"weights_gb":238,"kv_cache_gb":6.04,"overhead_gb":10.02,"total_gb":254.06,"total_gb_q8_cache":251.04},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":4096,"weights_gb":126.14,"kv_cache_gb":0.09,"overhead_gb":5.55,"total_gb":131.78,"total_gb_q8_cache":131.73},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":8192,"weights_gb":126.14,"kv_cache_gb":0.19,"overhead_gb":5.55,"total_gb":131.87,"total_gb_q8_cache":131.78},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":16384,"weights_gb":126.14,"kv_cache_gb":0.38,"overhead_gb":5.55,"total_gb":132.06,"total_gb_q8_cache":131.87},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":32768,"weights_gb":126.14,"kv_cache_gb":0.75,"overhead_gb":5.55,"total_gb":132.44,"total_gb_q8_cache":132.06},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":65536,"weights_gb":126.14,"kv_cache_gb":1.51,"overhead_gb":5.55,"total_gb":133.2,"total_gb_q8_cache":132.44},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":131072,"weights_gb":126.14,"kv_cache_gb":3.02,"overhead_gb":5.55,"total_gb":134.71,"total_gb_q8_cache":133.2},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q8_0","bytes_per_param":1.06,"context_tokens":262144,"weights_gb":126.14,"kv_cache_gb":6.04,"overhead_gb":5.55,"total_gb":137.73,"total_gb_q8_cache":134.71},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":4096,"weights_gb":97.58,"kv_cache_gb":0.09,"overhead_gb":4.4,"total_gb":102.08,"total_gb_q8_cache":102.03},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":8192,"weights_gb":97.58,"kv_cache_gb":0.19,"overhead_gb":4.4,"total_gb":102.17,"total_gb_q8_cache":102.08},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":16384,"weights_gb":97.58,"kv_cache_gb":0.38,"overhead_gb":4.4,"total_gb":102.36,"total_gb_q8_cache":102.17},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":32768,"weights_gb":97.58,"kv_cache_gb":0.75,"overhead_gb":4.4,"total_gb":102.74,"total_gb_q8_cache":102.36},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":65536,"weights_gb":97.58,"kv_cache_gb":1.51,"overhead_gb":4.4,"total_gb":103.49,"total_gb_q8_cache":102.74},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":131072,"weights_gb":97.58,"kv_cache_gb":3.02,"overhead_gb":4.4,"total_gb":105,"total_gb_q8_cache":103.49},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q6_K","bytes_per_param":0.82,"context_tokens":262144,"weights_gb":97.58,"kv_cache_gb":6.04,"overhead_gb":4.4,"total_gb":108.02,"total_gb_q8_cache":105},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":4096,"weights_gb":84.49,"kv_cache_gb":0.09,"overhead_gb":3.88,"total_gb":88.46,"total_gb_q8_cache":88.42},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":8192,"weights_gb":84.49,"kv_cache_gb":0.19,"overhead_gb":3.88,"total_gb":88.56,"total_gb_q8_cache":88.46},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":16384,"weights_gb":84.49,"kv_cache_gb":0.38,"overhead_gb":3.88,"total_gb":88.75,"total_gb_q8_cache":88.56},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":32768,"weights_gb":84.49,"kv_cache_gb":0.75,"overhead_gb":3.88,"total_gb":89.12,"total_gb_q8_cache":88.75},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":65536,"weights_gb":84.49,"kv_cache_gb":1.51,"overhead_gb":3.88,"total_gb":89.88,"total_gb_q8_cache":89.12},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":131072,"weights_gb":84.49,"kv_cache_gb":3.02,"overhead_gb":3.88,"total_gb":91.39,"total_gb_q8_cache":89.88},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q5_K_M","bytes_per_param":0.71,"context_tokens":262144,"weights_gb":84.49,"kv_cache_gb":6.04,"overhead_gb":3.88,"total_gb":94.41,"total_gb_q8_cache":91.39},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":4096,"weights_gb":69.02,"kv_cache_gb":0.09,"overhead_gb":3.26,"total_gb":72.38,"total_gb_q8_cache":72.33},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":8192,"weights_gb":69.02,"kv_cache_gb":0.19,"overhead_gb":3.26,"total_gb":72.47,"total_gb_q8_cache":72.38},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":16384,"weights_gb":69.02,"kv_cache_gb":0.38,"overhead_gb":3.26,"total_gb":72.66,"total_gb_q8_cache":72.47},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":32768,"weights_gb":69.02,"kv_cache_gb":0.75,"overhead_gb":3.26,"total_gb":73.04,"total_gb_q8_cache":72.66},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":65536,"weights_gb":69.02,"kv_cache_gb":1.51,"overhead_gb":3.26,"total_gb":73.79,"total_gb_q8_cache":73.04},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":131072,"weights_gb":69.02,"kv_cache_gb":3.02,"overhead_gb":3.26,"total_gb":75.3,"total_gb_q8_cache":73.79},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q4_K_M","bytes_per_param":0.58,"context_tokens":262144,"weights_gb":69.02,"kv_cache_gb":6.04,"overhead_gb":3.26,"total_gb":78.32,"total_gb_q8_cache":75.3},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":4096,"weights_gb":55.93,"kv_cache_gb":0.09,"overhead_gb":2.74,"total_gb":58.76,"total_gb_q8_cache":58.71},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":8192,"weights_gb":55.93,"kv_cache_gb":0.19,"overhead_gb":2.74,"total_gb":58.86,"total_gb_q8_cache":58.76},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":16384,"weights_gb":55.93,"kv_cache_gb":0.38,"overhead_gb":2.74,"total_gb":59.04,"total_gb_q8_cache":58.86},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":32768,"weights_gb":55.93,"kv_cache_gb":0.75,"overhead_gb":2.74,"total_gb":59.42,"total_gb_q8_cache":59.04},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":65536,"weights_gb":55.93,"kv_cache_gb":1.51,"overhead_gb":2.74,"total_gb":60.18,"total_gb_q8_cache":59.42},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":131072,"weights_gb":55.93,"kv_cache_gb":3.02,"overhead_gb":2.74,"total_gb":61.69,"total_gb_q8_cache":60.18},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","quant":"Q3_K_M","bytes_per_param":0.47,"context_tokens":262144,"weights_gb":55.93,"kv_cache_gb":6.04,"overhead_gb":2.74,"total_gb":64.71,"total_gb_q8_cache":61.69}
  ],
  "gpu_fit": [
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":36864,"tokens_per_second_ceiling":39},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":63488,"tokens_per_second_ceiling":74},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":70656,"tokens_per_second_ceiling":96},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":73728,"tokens_per_second_ceiling":111},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":77824,"tokens_per_second_ceiling":136},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":80896,"tokens_per_second_ceiling":168},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":69632,"tokens_per_second_ceiling":31},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":97280,"tokens_per_second_ceiling":59},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":103424,"tokens_per_second_ceiling":77},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":107520,"tokens_per_second_ceiling":89},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":110592,"tokens_per_second_ceiling":109},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":113664,"tokens_per_second_ceiling":134},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":36864,"tokens_per_second_ceiling":55},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":63488,"tokens_per_second_ceiling":104},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":70656,"tokens_per_second_ceiling":134},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":73728,"tokens_per_second_ceiling":155},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":77824,"tokens_per_second_ceiling":190},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":80896,"tokens_per_second_ceiling":235},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":102},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":193},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":250},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":288},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":353},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":436},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":110},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":208},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":269},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":311},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":380},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":469},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":196},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":370},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":478},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":552},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":676},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":834},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":105},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":198},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":256},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":296},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":362},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":447},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":94},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":178},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":230},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":266},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":326},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":402},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":196},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":370},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":478},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":552},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":676},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":834},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":223},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":421},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":544},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":628},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":769},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":949},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":366},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":691},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":894},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1032},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1263},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1559},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":88},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":165},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":213},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":246},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":302},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":372},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":60},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":113},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":146},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":168},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":206},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":254},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":69632,"tokens_per_second_ceiling":49},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":97280,"tokens_per_second_ceiling":92},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":103424,"tokens_per_second_ceiling":120},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":107520,"tokens_per_second_ceiling":138},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":110592,"tokens_per_second_ceiling":169},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":113664,"tokens_per_second_ceiling":209},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":105},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":198},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":256},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":296},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":362},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":447},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":30},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":56},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":73},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":84},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":103},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":127},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":28},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":53},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":68},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":79},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":97},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":119},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":67},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":127},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":164},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":189},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":232},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":286},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":90},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":169},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":218},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":252},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":309},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":381},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":8.1,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":131},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":4.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":248},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":4.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":320},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":3.8,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":370},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":453},
    {"model_id":"llama-3.2-3b","model":"Llama 3.2 3B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":559},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"tight","max_context_tokens":32768,"tokens_per_second_ceiling":27},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":51},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":65},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":76},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":92},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":114},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":21},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":40},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":52},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":60},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":74},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":91},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"tight","max_context_tokens":32768,"tokens_per_second_ceiling":38},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":71},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":92},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":106},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":129},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":160},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":70},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":132},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":170},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":196},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":240},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":297},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":75},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":142},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":183},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":211},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":259},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":319},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":133},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":252},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":325},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":376},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":460},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":568},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":71},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":135},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":174},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":201},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":247},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":304},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":64},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":121},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":157},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":181},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":222},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":274},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":133},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":252},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":325},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":376},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":460},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":568},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":152},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":286},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":370},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":428},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":524},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":646},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":249},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":471},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":608},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":703},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":860},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1062},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":60},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":112},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":145},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":168},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":205},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":254},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":41},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":77},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":99},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":115},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":140},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":173},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":33},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":63},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":81},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":94},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":115},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":142},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":71},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":135},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":174},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":201},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":247},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":304},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":20},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":38},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":50},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":57},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":70},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":87},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":19},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":36},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":46},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":54},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":66},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":81},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":46},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":86},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":112},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":129},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":158},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":195},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":61},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":115},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":149},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":172},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":210},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":260},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":10.59,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":89},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":169},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":4.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":218},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":4.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":252},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":3.65,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":308},
    {"model_id":"qwen3.5-4b","model":"Qwen3.5 4B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":3.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":380},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"tight","max_context_tokens":15360,"tokens_per_second_ceiling":30},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":30720,"tokens_per_second_ceiling":38},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":37888,"tokens_per_second_ceiling":44},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":46080,"tokens_per_second_ceiling":54},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":53248,"tokens_per_second_ceiling":67},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":44032,"tokens_per_second_ceiling":24},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":59392,"tokens_per_second_ceiling":31},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":66560,"tokens_per_second_ceiling":35},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":74752,"tokens_per_second_ceiling":43},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":81920,"tokens_per_second_ceiling":54},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"tight","max_context_tokens":15360,"tokens_per_second_ceiling":42},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":30720,"tokens_per_second_ceiling":54},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":37888,"tokens_per_second_ceiling":62},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":46080,"tokens_per_second_ceiling":76},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":53248,"tokens_per_second_ceiling":94},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"fits","max_context_tokens":43008,"tokens_per_second_ceiling":41},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":102400,"tokens_per_second_ceiling":77},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":117760,"tokens_per_second_ceiling":100},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":124928,"tokens_per_second_ceiling":115},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":141},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":174},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"fits","max_context_tokens":43008,"tokens_per_second_ceiling":44},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":102400,"tokens_per_second_ceiling":83},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":117760,"tokens_per_second_ceiling":108},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":124928,"tokens_per_second_ceiling":124},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":152},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":188},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"fits","max_context_tokens":100352,"tokens_per_second_ceiling":78},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":148},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":191},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":221},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":270},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":334},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":42},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":79},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":102},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":118},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":145},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":179},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":38},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":71},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":92},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":106},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":130},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":161},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":78},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":148},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":191},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":221},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":270},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":334},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":89},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":168},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":218},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":251},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":308},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":380},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":147},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":277},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":357},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":413},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":505},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":624},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":35},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":66},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":85},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":99},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":121},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":149},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":24},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":45},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":58},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":67},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":82},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":102},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":44032,"tokens_per_second_ceiling":37},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":59392,"tokens_per_second_ceiling":48},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":66560,"tokens_per_second_ceiling":55},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":74752,"tokens_per_second_ceiling":68},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":81920,"tokens_per_second_ceiling":83},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"fits","max_context_tokens":43008,"tokens_per_second_ceiling":42},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":102400,"tokens_per_second_ceiling":79},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":117760,"tokens_per_second_ceiling":102},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":124928,"tokens_per_second_ceiling":118},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":145},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":179},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":12},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":23},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":29},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":34},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":41},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":51},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":11},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":21},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":27},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":32},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":39},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":48},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":27},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":51},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":66},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":76},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":93},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":114},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":36},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":68},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":87},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":101},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":124},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":152},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":18.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":53},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":10.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":99},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":8.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":128},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":148},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.4,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":181},
    {"model_id":"llama-3.1-8b","model":"Llama 3.1 8B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.48,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":223},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"tight","max_context_tokens":12288,"tokens_per_second_ceiling":29},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":25600,"tokens_per_second_ceiling":37},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":43},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":39936,"tokens_per_second_ceiling":53},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":65},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":37888,"tokens_per_second_ceiling":23},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":30},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":35},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":42},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":52},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"tight","max_context_tokens":12288,"tokens_per_second_ceiling":41},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":25600,"tokens_per_second_ceiling":52},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":61},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":39936,"tokens_per_second_ceiling":74},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":92},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"fits","max_context_tokens":34816,"tokens_per_second_ceiling":40},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":75},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":97},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":113},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":138},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":170},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"fits","max_context_tokens":34816,"tokens_per_second_ceiling":43},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":81},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":105},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":121},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":148},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":183},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":76},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":144},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":187},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":215},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":264},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":325},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":41},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":77},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":100},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":115},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":141},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":174},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":37},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":70},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":90},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":104},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":127},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":157},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":76},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":144},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":187},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":215},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":264},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":325},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":87},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":164},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":212},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":245},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":300},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":370},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":143},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":270},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":349},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":403},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":493},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":608},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":34},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":64},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":83},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":96},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":118},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":145},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":23},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":44},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":57},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":66},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":80},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":99},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":37888,"tokens_per_second_ceiling":36},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":47},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":54},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":66},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":81},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"fits","max_context_tokens":34816,"tokens_per_second_ceiling":41},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":77},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":100},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":115},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":141},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":174},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":12},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":22},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":28},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":33},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":40},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":50},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":11},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":21},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":27},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":31},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":38},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":46},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":26},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":49},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":64},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":74},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":90},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":112},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":35},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":66},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":85},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":98},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":121},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":149},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":18.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":51},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":10.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":97},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":8.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":125},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.76,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":144},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.65,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":177},
    {"model_id":"qwen3-8b","model":"Qwen3 8B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.72,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":218},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"no","max_context_tokens":4096,"tokens_per_second_ceiling":25},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":77824,"tokens_per_second_ceiling":32},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":111616,"tokens_per_second_ceiling":37},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":45},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":55},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":119808,"tokens_per_second_ceiling":20},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":25},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":29},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":36},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":44},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"no","max_context_tokens":4096,"tokens_per_second_ceiling":34},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":77824,"tokens_per_second_ceiling":44},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":111616,"tokens_per_second_ceiling":51},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":63},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":77},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"tight","max_context_tokens":62464,"tokens_per_second_ceiling":34},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":64},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":82},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":95},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":116},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":144},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"tight","max_context_tokens":62464,"tokens_per_second_ceiling":36},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":69},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":89},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":102},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":125},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":155},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":65},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":122},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":158},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":182},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":223},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":275},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":35},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":65},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":84},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":98},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":119},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":147},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":31},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":59},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":76},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":88},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":108},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":133},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":65},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":122},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":158},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":182},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":223},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":275},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":74},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":139},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":179},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":207},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":254},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":313},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":121},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":228},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":295},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":340},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":417},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":514},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":29},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":54},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":70},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":81},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":100},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":123},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":20},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":37},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":48},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":55},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":68},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":84},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":119808,"tokens_per_second_ceiling":30},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":39},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":46},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":56},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":69},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"tight","max_context_tokens":62464,"tokens_per_second_ceiling":35},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":65},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":84},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":98},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":119},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":147},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":10},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":19},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":24},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":28},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":34},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":42},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":9},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":23},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":26},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":32},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":39},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":22},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":42},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":54},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":62},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":76},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":94},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":30},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":56},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":72},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":83},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":102},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":126},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":20.99,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":43},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":11.51,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":82},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":9.09,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":106},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":7.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":122},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":6.67,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":149},
    {"model_id":"qwen3.5-9b","model":"Qwen3.5 9B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":5.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":184},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"tight","max_context_tokens":19456,"tokens_per_second_ceiling":26},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":103424,"tokens_per_second_ceiling":30},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":36},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":45},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"tight","max_context_tokens":68608,"tokens_per_second_ceiling":16},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":20},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":24},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":29},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":36},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"tight","max_context_tokens":19456,"tokens_per_second_ceiling":36},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":103424,"tokens_per_second_ceiling":41},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":51},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":63},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":52},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":67},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":77},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":94},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":116},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":55},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":72},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":83},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":101},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":125},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":52},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":99},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":127},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":147},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":180},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":222},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":28},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":53},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":68},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":79},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":97},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":119},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":25},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":48},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":61},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":71},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":87},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":107},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":52},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":99},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":127},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":147},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":180},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":222},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":59},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":112},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":145},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":168},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":205},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":253},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":98},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":184},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":238},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":275},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":337},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":416},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":23},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":44},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":57},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":66},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":80},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":99},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":16},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":30},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":39},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":45},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":55},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":68},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"tight","max_context_tokens":68608,"tokens_per_second_ceiling":25},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":32},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":37},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":45},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":56},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":53},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":68},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":79},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":97},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":119},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":8},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":15},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":19},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":22},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":27},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":34},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":7},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":14},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":18},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":21},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":26},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":32},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":18},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":34},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":44},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":50},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":62},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":76},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":24},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":45},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":58},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":67},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":82},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":102},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":25.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":35},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":14.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":66},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":11.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":85},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":9.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":99},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":121},
    {"model_id":"gemma-4-12b","model":"Gemma 4 12B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":6.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":149},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"no","max_context_tokens":2048,"tokens_per_second_ceiling":25},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"tight","max_context_tokens":23552,"tokens_per_second_ceiling":29},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":48128,"tokens_per_second_ceiling":36},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":69632,"tokens_per_second_ceiling":44},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"tight","max_context_tokens":13312,"tokens_per_second_ceiling":16},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":60416,"tokens_per_second_ceiling":20},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":80896,"tokens_per_second_ceiling":23},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":106496,"tokens_per_second_ceiling":28},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":128000,"tokens_per_second_ceiling":35},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"no","max_context_tokens":2048,"tokens_per_second_ceiling":35},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"tight","max_context_tokens":23552,"tokens_per_second_ceiling":41},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":48128,"tokens_per_second_ceiling":50},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":69632,"tokens_per_second_ceiling":62},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"fits","max_context_tokens":129024,"tokens_per_second_ceiling":51},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":65},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":76},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":93},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":114},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"fits","max_context_tokens":129024,"tokens_per_second_ceiling":55},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":71},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":81},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":100},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":123},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"fits","max_context_tokens":63488,"tokens_per_second_ceiling":51},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":97},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":125},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":145},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":177},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":219},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":28},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":52},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":67},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":78},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":95},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":117},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":25},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":47},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":60},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":70},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":85},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":105},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":51},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":97},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":125},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":145},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":177},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":219},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":58},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":110},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":143},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":165},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":202},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":249},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":96},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":181},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":234},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":271},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":331},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":409},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":23},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":43},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":56},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":65},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":79},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":98},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":16},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":30},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":38},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":44},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":54},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":67},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"tight","max_context_tokens":13312,"tokens_per_second_ceiling":24},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":60416,"tokens_per_second_ceiling":31},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":80896,"tokens_per_second_ceiling":36},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":106496,"tokens_per_second_ceiling":44},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":128000,"tokens_per_second_ceiling":55},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"fits","max_context_tokens":129024,"tokens_per_second_ceiling":52},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":67},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":78},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":95},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":117},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":8},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":15},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":19},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":22},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":27},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":33},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":7},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":14},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":18},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":21},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":25},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":31},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":18},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":33},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":43},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":50},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":61},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":75},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":23},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":44},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":57},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":66},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":81},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":100},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":26.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":34},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":14.82,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":65},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":11.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":84},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":10.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":97},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":8.73,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":119},
    {"model_id":"gemma-3-12b","model":"Gemma 3 12B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":7.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":146},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"no","max_context_tokens":3072,"tokens_per_second_ceiling":25},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"tight","max_context_tokens":14336,"tokens_per_second_ceiling":31},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":24576,"tokens_per_second_ceiling":38},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"tight","max_context_tokens":16384,"tokens_per_second_ceiling":18},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":26624,"tokens_per_second_ceiling":20},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":37888,"tokens_per_second_ceiling":25},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":47104,"tokens_per_second_ceiling":31},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"no","max_context_tokens":3072,"tokens_per_second_ceiling":35},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"tight","max_context_tokens":14336,"tokens_per_second_ceiling":43},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":24576,"tokens_per_second_ceiling":54},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":44},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"fits","max_context_tokens":62464,"tokens_per_second_ceiling":57},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":72704,"tokens_per_second_ceiling":66},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":83968,"tokens_per_second_ceiling":81},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":94208,"tokens_per_second_ceiling":100},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":48},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"fits","max_context_tokens":62464,"tokens_per_second_ceiling":61},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":72704,"tokens_per_second_ceiling":71},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":83968,"tokens_per_second_ceiling":87},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":94208,"tokens_per_second_ceiling":107},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"no","max_context_tokens":4096,"tokens_per_second_ceiling":45},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"fits","max_context_tokens":88064,"tokens_per_second_ceiling":85},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"fits","max_context_tokens":109568,"tokens_per_second_ceiling":109},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":118784,"tokens_per_second_ceiling":126},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":130048,"tokens_per_second_ceiling":154},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":191},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"fits","max_context_tokens":97280,"tokens_per_second_ceiling":24},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":45},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":59},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":68},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":83},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":102},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"fits","max_context_tokens":97280,"tokens_per_second_ceiling":22},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":41},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":53},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":61},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":74},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":92},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":45},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":85},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":109},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":126},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":154},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":191},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":51},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":96},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":124},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":144},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":176},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":217},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":84},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":158},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":204},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":236},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":289},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":356},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":20},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":38},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":49},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":56},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":69},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":85},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":14},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":26},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":33},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":38},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":47},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":58},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"tight","max_context_tokens":16384,"tokens_per_second_ceiling":27},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":26624,"tokens_per_second_ceiling":32},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":37888,"tokens_per_second_ceiling":39},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":47104,"tokens_per_second_ceiling":48},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":45},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"fits","max_context_tokens":62464,"tokens_per_second_ceiling":59},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":72704,"tokens_per_second_ceiling":68},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":83968,"tokens_per_second_ceiling":83},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":94208,"tokens_per_second_ceiling":102},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":7},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":13},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":19},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":24},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":29},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":6},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":12},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":16},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":18},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":22},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":27},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":15},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":29},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":37},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":43},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":53},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":65},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":20},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":39},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":50},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":58},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":71},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":87},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":30.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":30},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":17.28,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":57},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":13.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":73},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.18,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":85},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.29,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":103},
    {"model_id":"ministral-3-14b","model":"Ministral 3 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":8.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":128},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"tight","max_context_tokens":11264,"tokens_per_second_ceiling":29},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":21504,"tokens_per_second_ceiling":36},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"tight","max_context_tokens":12288,"tokens_per_second_ceiling":17},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":22528,"tokens_per_second_ceiling":19},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":34816,"tokens_per_second_ceiling":23},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":29},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"tight","max_context_tokens":11264,"tokens_per_second_ceiling":41},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":21504,"tokens_per_second_ceiling":51},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"fits","max_context_tokens":35840,"tokens_per_second_ceiling":42},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":54},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":62},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":76},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":94},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"fits","max_context_tokens":35840,"tokens_per_second_ceiling":45},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":58},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":67},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":82},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":101},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":80},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":103},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":119},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":146},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":180},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":23},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":43},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":55},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":64},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":78},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":97},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":20},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":39},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":50},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":58},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":70},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":87},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":42},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":80},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":103},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":119},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":146},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":180},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":48},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":91},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":118},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":136},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":166},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":205},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":79},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":149},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":193},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":223},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":273},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":337},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":19},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":36},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":46},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":53},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":65},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":81},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":13},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":24},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":31},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":36},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":45},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":55},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"tight","max_context_tokens":12288,"tokens_per_second_ceiling":26},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":22528,"tokens_per_second_ceiling":30},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":34816,"tokens_per_second_ceiling":37},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":45},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"fits","max_context_tokens":35840,"tokens_per_second_ceiling":43},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":55},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":64},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":78},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":97},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":6},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":12},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":16},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":18},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":22},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":27},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":6},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":11},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":15},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":17},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":21},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":26},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":15},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":27},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":35},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":41},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":50},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":62},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":19},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":37},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":47},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":55},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":67},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":82},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":32.63,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":28},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":18.16,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":54},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":14.46,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":69},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":12.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":80},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":10.77,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":98},
    {"model_id":"qwen3-14b","model":"Qwen3 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.08,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":121},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"tight","max_context_tokens":9216,"tokens_per_second_ceiling":30},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":36},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"tight","max_context_tokens":10240,"tokens_per_second_ceiling":17},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":19},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":24},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":29},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"tight","max_context_tokens":9216,"tokens_per_second_ceiling":41},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":51},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":42},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":54},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":63},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":77},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":95},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":45},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":59},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":68},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":83},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":102},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":81},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":104},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":120},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":147},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":182},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":23},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":43},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":56},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":64},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":79},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":97},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":21},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":39},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":50},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":58},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":71},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":88},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":43},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":81},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":104},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":120},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":147},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":182},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":49},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":92},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":118},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":137},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":167},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":207},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":80},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":150},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":195},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":225},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":275},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":339},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":19},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":36},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":46},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":54},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":66},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":81},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":13},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":25},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":32},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":37},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":45},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":55},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"tight","max_context_tokens":10240,"tokens_per_second_ceiling":26},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":30},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":37},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":45},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":43},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":56},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":64},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":79},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":97},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":7},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":12},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":16},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":18},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":22},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":28},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":6},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":12},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":15},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":17},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":21},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":26},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":15},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":28},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":36},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":41},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":50},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":62},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":20},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":37},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":48},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":55},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":67},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":83},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":32.75,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":29},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":18.38,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":54},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":14.71,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":70},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":13.03,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":80},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":11.04,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":99},
    {"model_id":"phi-4","model":"Phi-4 14B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":9.36,"verdict":"fits","max_context_tokens":16384,"tokens_per_second_ceiling":122},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"tight","max_context_tokens":25600,"tokens_per_second_ceiling":149},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":81920,"tokens_per_second_ceiling":97},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":119},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"tight","max_context_tokens":25600,"tokens_per_second_ceiling":209},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":222},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":256},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":314},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":387},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":239},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":276},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":338},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":417},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":329},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":425},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":491},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":601},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":741},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"tight","max_context_tokens":57344,"tokens_per_second_ceiling":93},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":176},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":228},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":263},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":322},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":397},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"tight","max_context_tokens":57344,"tokens_per_second_ceiling":84},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":158},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":205},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":237},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":290},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":357},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":174},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":329},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":425},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":491},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":601},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":741},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":198},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":374},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":484},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":558},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":684},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":844},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":326},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":615},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":794},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":917},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1123},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1386},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":78},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":147},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":190},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":219},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":268},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":331},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":53},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":100},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":129},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":150},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":183},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":226},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":81920,"tokens_per_second_ceiling":150},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":185},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":228},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":263},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":322},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":397},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":27},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":50},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":65},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":75},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":92},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":113},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":25},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":47},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":61},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":70},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":86},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":106},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":60},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":113},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":146},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":168},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":206},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":254},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":80},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":150},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":194},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":224},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":275},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":339},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":44.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":117},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":23.85,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":220},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":18.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":285},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":329},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":13.37,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":402},
    {"model_id":"gpt-oss-20b","model":"gpt-oss 20B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":10.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":496},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"no","max_context_tokens":1024,"tokens_per_second_ceiling":14},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":17408,"tokens_per_second_ceiling":18},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"tight","max_context_tokens":10240,"tokens_per_second_ceiling":33},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":27648,"tokens_per_second_ceiling":38},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":47104,"tokens_per_second_ceiling":47},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":63488,"tokens_per_second_ceiling":58},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"tight","max_context_tokens":10240,"tokens_per_second_ceiling":36},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":27648,"tokens_per_second_ceiling":41},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":47104,"tokens_per_second_ceiling":51},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":63488,"tokens_per_second_ceiling":63},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"tight","max_context_tokens":20480,"tokens_per_second_ceiling":49},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":57344,"tokens_per_second_ceiling":64},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":73728,"tokens_per_second_ceiling":74},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":93184,"tokens_per_second_ceiling":90},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":110592,"tokens_per_second_ceiling":111},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":113664,"tokens_per_second_ceiling":26},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":34},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":39},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":48},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":60},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":113664,"tokens_per_second_ceiling":24},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":31},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":35},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":43},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":54},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":26},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":49},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":64},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":74},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":90},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":111},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":30},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":56},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":73},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":84},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":103},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":127},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":49},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":92},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":119},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":138},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":168},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":208},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":12},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":22},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":28},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":33},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":40},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":50},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":8},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":15},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":19},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":22},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":27},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":34},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"no","max_context_tokens":1024,"tokens_per_second_ceiling":23},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":17408,"tokens_per_second_ceiling":28},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"tight","max_context_tokens":10240,"tokens_per_second_ceiling":34},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":27648,"tokens_per_second_ceiling":39},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":47104,"tokens_per_second_ceiling":48},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":63488,"tokens_per_second_ceiling":60},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":4},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":8},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":10},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":11},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":14},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":4},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":7},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":9},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":11},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":13},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":16},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":9},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":22},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":25},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":31},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":38},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":12},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":23},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":29},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":34},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":41},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":51},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":18},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":33},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":43},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":49},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":60},
    {"model_id":"devstral-small-2-24b","model":"Devstral Small 2 24B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":74},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"no","max_context_tokens":1024,"tokens_per_second_ceiling":14},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":17408,"tokens_per_second_ceiling":18},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"tight","max_context_tokens":10240,"tokens_per_second_ceiling":33},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":27648,"tokens_per_second_ceiling":38},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":47104,"tokens_per_second_ceiling":47},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":63488,"tokens_per_second_ceiling":58},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"tight","max_context_tokens":10240,"tokens_per_second_ceiling":36},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":27648,"tokens_per_second_ceiling":41},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":47104,"tokens_per_second_ceiling":51},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":63488,"tokens_per_second_ceiling":63},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"tight","max_context_tokens":20480,"tokens_per_second_ceiling":49},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":57344,"tokens_per_second_ceiling":64},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":73728,"tokens_per_second_ceiling":74},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":93184,"tokens_per_second_ceiling":90},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":110592,"tokens_per_second_ceiling":111},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":113664,"tokens_per_second_ceiling":26},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":34},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":39},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":48},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":60},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":113664,"tokens_per_second_ceiling":24},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":31},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":35},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":43},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":54},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":26},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":49},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":64},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":74},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":90},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":111},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":30},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":56},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":73},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":84},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":103},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":127},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":49},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":92},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":119},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":138},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":168},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":208},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":12},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":22},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":28},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":33},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":40},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":50},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":8},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":15},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":19},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":22},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":27},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":34},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"no","max_context_tokens":1024,"tokens_per_second_ceiling":23},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":17408,"tokens_per_second_ceiling":28},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"tight","max_context_tokens":10240,"tokens_per_second_ceiling":34},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":27648,"tokens_per_second_ceiling":39},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":47104,"tokens_per_second_ceiling":48},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":63488,"tokens_per_second_ceiling":60},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":4},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":8},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":10},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":11},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":14},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":4},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":7},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":9},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":11},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":13},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":16},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":9},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":22},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":25},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":31},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":38},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":12},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":23},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":29},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":34},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":41},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":51},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":51.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":18},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":28.3,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":33},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":43},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":49},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":60},
    {"model_id":"mistral-small-3.1-24b","model":"Mistral Small 3.1 24B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.57,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":74},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":91136,"tokens_per_second_ceiling":113},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"no","max_context_tokens":4096,"tokens_per_second_ceiling":210},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":243},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":297},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":367},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"no","max_context_tokens":4096,"tokens_per_second_ceiling":226},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":262},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":320},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":395},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"tight","max_context_tokens":60416,"tokens_per_second_ceiling":311},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":403},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":465},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":569},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":702},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":167},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":216},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":249},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":305},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":376},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":150},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":194},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":224},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":274},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":339},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":165},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":311},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":403},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":465},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":569},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":702},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":188},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":354},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":458},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":529},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":648},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":799},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":309},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":582},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":753},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":869},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1064},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1313},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":74},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":139},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":180},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":208},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":254},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":314},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":50},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":95},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":123},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":142},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":173},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":214},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":91136,"tokens_per_second_ceiling":176},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"no","max_context_tokens":4096,"tokens_per_second_ceiling":216},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":249},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":305},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":376},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":25},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":47},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":61},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":71},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":87},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":107},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":24},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":44},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":58},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":66},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":81},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":100},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":57},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":107},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":138},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":159},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":195},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":241},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":75},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":142},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":184},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":212},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":260},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":321},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":54.54,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":111},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":29.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":209},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":22.88,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":270},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":19.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":311},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":16.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":381},
    {"model_id":"gemma-4-26b-a4b","model":"Gemma 4 26B-A4B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":13.49,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":470},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"tight","max_context_tokens":10240,"tokens_per_second_ceiling":16},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"tight","max_context_tokens":19456,"tokens_per_second_ceiling":34},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"fits","max_context_tokens":64512,"tokens_per_second_ceiling":41},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"fits","max_context_tokens":102400,"tokens_per_second_ceiling":51},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"tight","max_context_tokens":19456,"tokens_per_second_ceiling":36},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"fits","max_context_tokens":64512,"tokens_per_second_ceiling":44},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"fits","max_context_tokens":102400,"tokens_per_second_ceiling":55},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"fits","max_context_tokens":73728,"tokens_per_second_ceiling":56},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"fits","max_context_tokens":112640,"tokens_per_second_ceiling":64},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":79},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":97},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":23},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":30},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":35},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":42},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":52},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":21},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":27},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":31},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":38},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":47},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":23},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":43},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":56},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":64},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":79},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":97},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":26},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":49},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":64},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":73},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":90},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":111},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":43},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":81},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":104},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":121},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":148},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":182},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":10},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":19},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":25},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":29},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":35},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":43},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":7},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":13},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":20},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":24},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":30},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"tight","max_context_tokens":10240,"tokens_per_second_ceiling":24},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"tight","max_context_tokens":19456,"tokens_per_second_ceiling":35},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"fits","max_context_tokens":64512,"tokens_per_second_ceiling":42},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"fits","max_context_tokens":102400,"tokens_per_second_ceiling":52},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":3},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":7},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":9},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":10},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":12},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":15},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":3},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":6},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":8},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":9},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":11},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":14},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":8},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":15},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":19},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":22},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":27},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":33},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":10},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":20},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":26},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":29},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":36},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":45},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":58.6,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":15},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":31.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":29},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":24.97,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":37},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":43},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":18.13,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":53},
    {"model_id":"gemma-3-27b","model":"Gemma 3 27B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":15,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":65},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"tight","max_context_tokens":14336,"tokens_per_second_ceiling":15},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"tight","max_context_tokens":24576,"tokens_per_second_ceiling":33},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"fits","max_context_tokens":81920,"tokens_per_second_ceiling":41},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"fits","max_context_tokens":130048,"tokens_per_second_ceiling":50},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"tight","max_context_tokens":24576,"tokens_per_second_ceiling":36},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"fits","max_context_tokens":81920,"tokens_per_second_ceiling":44},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"fits","max_context_tokens":130048,"tokens_per_second_ceiling":54},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"fits","max_context_tokens":92160,"tokens_per_second_ceiling":55},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":64},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":78},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":96},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":23},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":29},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":34},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":42},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":51},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":21},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":27},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":31},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":38},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":46},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":23},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":43},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":55},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":64},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":78},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":96},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":26},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":48},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":63},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":72},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":89},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":109},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":42},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":80},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":103},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":119},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":145},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":179},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":10},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":19},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":25},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":28},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":35},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":43},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":7},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":13},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":19},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":24},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":29},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"tight","max_context_tokens":14336,"tokens_per_second_ceiling":24},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"tight","max_context_tokens":24576,"tokens_per_second_ceiling":34},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"fits","max_context_tokens":81920,"tokens_per_second_ceiling":42},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"fits","max_context_tokens":130048,"tokens_per_second_ceiling":51},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":3},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":6},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":8},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":10},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":12},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":15},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":3},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":6},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":8},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":9},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":11},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":14},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":8},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":15},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":19},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":22},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":27},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":33},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":10},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":19},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":25},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":29},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":36},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":44},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":59.01,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":15},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":31.83,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":29},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":24.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":37},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":21.71,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":43},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":17.96,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":52},
    {"model_id":"qwen3.8-27b","model":"Qwen3.8 27B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":14.78,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":64},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"fits","max_context_tokens":38912,"tokens_per_second_ceiling":342},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":422},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"fits","max_context_tokens":38912,"tokens_per_second_ceiling":369},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":455},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"tight","max_context_tokens":38912,"tokens_per_second_ceiling":464},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":535},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":655},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":809},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":192},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":248},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":287},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":351},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":433},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":173},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":224},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":258},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":316},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":390},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":190},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":359},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":464},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":535},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":655},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":809},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":216},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":408},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":527},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":609},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":746},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":920},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":355},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":670},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":867},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":1001},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":1225},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":1512},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":85},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":160},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":207},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":239},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":293},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":361},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":58},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":109},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":141},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":163},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":200},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":246},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"fits","max_context_tokens":38912,"tokens_per_second_ceiling":351},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":433},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":29},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":55},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":71},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":82},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":100},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":123},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":27},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":51},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":66},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":76},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":94},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":116},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":65},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":123},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":159},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":183},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":225},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":277},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":87},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":164},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":212},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":245},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":300},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":370},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":64.75,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":127},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":34.93,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":240},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.32,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":310},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.83,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":359},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.7,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":439},
    {"model_id":"qwen3-30b-a3b","model":"Qwen3 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.21,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":542},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"fits","max_context_tokens":63488,"tokens_per_second_ceiling":377},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"fits","max_context_tokens":130048,"tokens_per_second_ceiling":465},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"fits","max_context_tokens":63488,"tokens_per_second_ceiling":406},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"fits","max_context_tokens":130048,"tokens_per_second_ceiling":500},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"tight","max_context_tokens":60416,"tokens_per_second_ceiling":510},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"fits","max_context_tokens":125952,"tokens_per_second_ceiling":589},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":721},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":890},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":211},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":273},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":315},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":386},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":477},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":190},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":246},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":284},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":348},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":429},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":209},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":394},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":510},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":589},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":721},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":890},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":238},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":449},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":580},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":670},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":820},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1012},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":391},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":737},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":953},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1101},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1348},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1663},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":93},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":176},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":228},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":263},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":322},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":397},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":64},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":120},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":155},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":179},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":220},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":271},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"fits","max_context_tokens":63488,"tokens_per_second_ceiling":386},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"fits","max_context_tokens":130048,"tokens_per_second_ceiling":477},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":32},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":60},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":78},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":90},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":110},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":136},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":30},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":56},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":73},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":84},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":103},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":127},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":72},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":135},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":175},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":202},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":247},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":305},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":96},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":180},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":233},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":269},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":329},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":407},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":65.84,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":140},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":35.34,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":264},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":341},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.98,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":394},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.76,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":483},
    {"model_id":"glm-4.7-flash","model":"GLM-4.7 Flash 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.19,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":596},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"tight","max_context_tokens":30720,"tokens_per_second_ceiling":36},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"fits","max_context_tokens":74752,"tokens_per_second_ceiling":45},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"tight","max_context_tokens":30720,"tokens_per_second_ceiling":39},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"fits","max_context_tokens":74752,"tokens_per_second_ceiling":48},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"tight","max_context_tokens":28672,"tokens_per_second_ceiling":49},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"fits","max_context_tokens":71680,"tokens_per_second_ceiling":56},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"fits","max_context_tokens":123904,"tokens_per_second_ceiling":69},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":85},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"fits","max_context_tokens":118784,"tokens_per_second_ceiling":20},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":26},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":30},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":37},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":46},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"fits","max_context_tokens":118784,"tokens_per_second_ceiling":18},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":24},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":27},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":33},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":41},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":20},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":38},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":49},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":56},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":69},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":85},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"fits","max_context_tokens":115712,"tokens_per_second_ceiling":23},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":43},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":56},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":64},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":79},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":97},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"fits","max_context_tokens":115712,"tokens_per_second_ceiling":37},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":71},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":91},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":106},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":129},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":159},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":9},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":22},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":25},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":31},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":38},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":6},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":12},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":15},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":21},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":26},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"tight","max_context_tokens":30720,"tokens_per_second_ceiling":37},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"fits","max_context_tokens":74752,"tokens_per_second_ceiling":46},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":3},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":6},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":7},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":9},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":11},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":13},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":3},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":5},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":7},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":8},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":10},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":12},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":7},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":13},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":19},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":24},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":29},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":9},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":22},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":26},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":32},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":39},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":67.11,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":13},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":36.52,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":25},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":28.7,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":33},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":25.12,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":38},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":20.89,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":46},
    {"model_id":"gemma-4-31b","model":"Gemma 4 31B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":17.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":57},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":377},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":465},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":406},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":500},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"tight","max_context_tokens":131072,"tokens_per_second_ceiling":510},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":589},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":721},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":890},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":211},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":273},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":315},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":386},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":477},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":190},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":246},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":284},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":348},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":429},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":209},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":394},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":510},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":589},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":721},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":890},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":238},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":449},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":580},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":670},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":820},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1012},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":391},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":737},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":953},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1101},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1348},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1663},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":93},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":176},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":228},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":263},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":322},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":397},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":64},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":120},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":155},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":179},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":220},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":271},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":386},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":477},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":32},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":60},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":78},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":90},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":110},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":136},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":30},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":56},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":73},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":84},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":103},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":127},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":72},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":135},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":175},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":202},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":247},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":305},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":96},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":180},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":233},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":269},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":329},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":407},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":66.33,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":140},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":35.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":264},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":27.55,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":341},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":23.93,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":394},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":19.66,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":483},
    {"model_id":"nemotron-3.5-lightning-30b-a3b","model":"Nemotron 3.5 Lightning 30B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":16.05,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":596},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"tight","max_context_tokens":9216,"tokens_per_second_ceiling":34},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":23552,"tokens_per_second_ceiling":43},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"tight","max_context_tokens":9216,"tokens_per_second_ceiling":37},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":23552,"tokens_per_second_ceiling":46},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":7168,"tokens_per_second_ceiling":47},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":21504,"tokens_per_second_ceiling":54},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":37888,"tokens_per_second_ceiling":66},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":81},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":33792,"tokens_per_second_ceiling":19},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":25},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":29},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":35},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":44},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":33792,"tokens_per_second_ceiling":17},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":22},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":26},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":32},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":39},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":19},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":36},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":47},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":54},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":66},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":81},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"tight","max_context_tokens":27648,"tokens_per_second_ceiling":22},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":41},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":53},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":61},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":75},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":93},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"tight","max_context_tokens":27648,"tokens_per_second_ceiling":36},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":67},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":87},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":101},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":123},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":152},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":9},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":16},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":21},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":24},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":29},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":36},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":6},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":11},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":14},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":16},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":20},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":25},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"tight","max_context_tokens":9216,"tokens_per_second_ceiling":35},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":23552,"tokens_per_second_ceiling":44},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":3},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":5},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":7},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":8},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":10},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":12},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":3},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":5},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":7},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":8},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":9},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":12},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":7},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":12},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":16},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":18},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":23},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":28},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":9},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":16},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":21},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":25},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":30},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":37},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":13},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":24},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":31},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":36},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":44},
    {"model_id":"qwen3-32b","model":"Qwen3 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":40960,"tokens_per_second_ceiling":54},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"tight","max_context_tokens":9216,"tokens_per_second_ceiling":34},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":23552,"tokens_per_second_ceiling":43},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"tight","max_context_tokens":9216,"tokens_per_second_ceiling":37},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":23552,"tokens_per_second_ceiling":46},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":7168,"tokens_per_second_ceiling":47},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":21504,"tokens_per_second_ceiling":54},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":66},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":81},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":19},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":25},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":29},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":35},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":44},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":17},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":22},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":26},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":32},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":39},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":19},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":36},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":47},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":54},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":66},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":81},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"tight","max_context_tokens":27648,"tokens_per_second_ceiling":22},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":41},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":53},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":61},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":75},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":93},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"tight","max_context_tokens":27648,"tokens_per_second_ceiling":36},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":67},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":87},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":101},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":123},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":152},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":9},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":16},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":21},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":24},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":29},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":36},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":6},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":11},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":14},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":16},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":20},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":25},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"tight","max_context_tokens":9216,"tokens_per_second_ceiling":35},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":23552,"tokens_per_second_ceiling":44},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":3},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":5},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":7},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":8},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":10},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":12},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":3},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":5},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":7},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":8},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":9},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":12},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":7},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":12},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":16},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":18},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":23},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":28},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":9},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":16},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":21},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":25},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":30},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":37},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":13},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":24},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":31},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":36},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":44},
    {"model_id":"qwen2.5-coder-32b","model":"Qwen2.5 Coder 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":32768,"tokens_per_second_ceiling":54},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"tight","max_context_tokens":9216,"tokens_per_second_ceiling":34},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":23552,"tokens_per_second_ceiling":43},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"tight","max_context_tokens":9216,"tokens_per_second_ceiling":37},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":23552,"tokens_per_second_ceiling":46},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":7168,"tokens_per_second_ceiling":47},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":21504,"tokens_per_second_ceiling":54},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":37888,"tokens_per_second_ceiling":66},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":52224,"tokens_per_second_ceiling":81},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":33792,"tokens_per_second_ceiling":19},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":64512,"tokens_per_second_ceiling":25},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":78848,"tokens_per_second_ceiling":29},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":96256,"tokens_per_second_ceiling":35},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":110592,"tokens_per_second_ceiling":44},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":33792,"tokens_per_second_ceiling":17},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":64512,"tokens_per_second_ceiling":22},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":78848,"tokens_per_second_ceiling":26},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":96256,"tokens_per_second_ceiling":32},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":110592,"tokens_per_second_ceiling":39},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":84992,"tokens_per_second_ceiling":19},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":36},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":47},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":54},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":66},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":81},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"tight","max_context_tokens":27648,"tokens_per_second_ceiling":22},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":41},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":53},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":61},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":75},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":93},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"tight","max_context_tokens":27648,"tokens_per_second_ceiling":36},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":67},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":87},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":101},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":123},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":152},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":9},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":16},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":21},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":24},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":29},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":36},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":84992,"tokens_per_second_ceiling":6},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":11},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":14},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":16},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":20},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":25},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"tight","max_context_tokens":9216,"tokens_per_second_ceiling":35},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":23552,"tokens_per_second_ceiling":44},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":3},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":5},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":7},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":8},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":10},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":12},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":84992,"tokens_per_second_ceiling":3},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":5},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":7},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":8},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":9},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":12},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":84992,"tokens_per_second_ceiling":7},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":12},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":16},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":18},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":23},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":28},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":9},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":16},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":21},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":25},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":30},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":37},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":70.87,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":13},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":38.81,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":24},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":30.62,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":31},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":26.87,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":36},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":44},
    {"model_id":"deepseek-r1-distill-qwen-32b","model":"DeepSeek-R1 Distill Qwen 32B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.68,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":54},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"tight","max_context_tokens":25600,"tokens_per_second_ceiling":377},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":465},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"tight","max_context_tokens":25600,"tokens_per_second_ceiling":406},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":500},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"tight","max_context_tokens":131072,"tokens_per_second_ceiling":589},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":721},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":890},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":211},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":273},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":315},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":386},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":477},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":190},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":246},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":284},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":348},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":429},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":209},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":394},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":510},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":589},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":721},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":890},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"tight","max_context_tokens":26624,"tokens_per_second_ceiling":238},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":449},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":580},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":670},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":820},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1012},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"tight","max_context_tokens":26624,"tokens_per_second_ceiling":391},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":737},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":953},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1101},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1348},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1663},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":93},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":176},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":228},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":263},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":322},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":397},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":64},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":120},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":155},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":179},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":220},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":271},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"tight","max_context_tokens":25600,"tokens_per_second_ceiling":386},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":477},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":32},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":60},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":78},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":90},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":110},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":136},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":30},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":56},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":73},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":84},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":103},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":127},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":72},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":135},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":175},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":202},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":247},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":305},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":96},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":180},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":233},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":269},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":329},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":407},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":75.61,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":140},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":40.41,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":264},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":31.43,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":341},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":27.31,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":394},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":22.44,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":483},
    {"model_id":"qwen3.6-35b-a3b","model":"Qwen3.6 35B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":18.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":596},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":7168,"tokens_per_second_ceiling":16},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":31744,"tokens_per_second_ceiling":20},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":7168,"tokens_per_second_ceiling":15},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":31744,"tokens_per_second_ceiling":18},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"fits","max_context_tokens":38912,"tokens_per_second_ceiling":17},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":92160,"tokens_per_second_ceiling":22},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":116736,"tokens_per_second_ceiling":25},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":31},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":38},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":46080,"tokens_per_second_ceiling":25},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":70656,"tokens_per_second_ceiling":28},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":100352,"tokens_per_second_ceiling":35},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":124928,"tokens_per_second_ceiling":43},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":46080,"tokens_per_second_ceiling":41},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":70656,"tokens_per_second_ceiling":47},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":100352,"tokens_per_second_ceiling":57},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":124928,"tokens_per_second_ceiling":71},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":7},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":10},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":11},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":14},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"fits","max_context_tokens":38912,"tokens_per_second_ceiling":5},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":92160,"tokens_per_second_ceiling":7},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":116736,"tokens_per_second_ceiling":8},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":9},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":12},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"fits","max_context_tokens":125952,"tokens_per_second_ceiling":3},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":3},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":4},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":5},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":6},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"fits","max_context_tokens":38912,"tokens_per_second_ceiling":2},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":92160,"tokens_per_second_ceiling":3},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":116736,"tokens_per_second_ceiling":4},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":4},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":5},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"fits","max_context_tokens":38912,"tokens_per_second_ceiling":6},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":92160,"tokens_per_second_ceiling":7},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":116736,"tokens_per_second_ceiling":9},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":10},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":13},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":4},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":8},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":10},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":11},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":14},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":6},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":11},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":15},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":21},
    {"model_id":"llama-3.3-70b","model":"Llama 3.3 70B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":25},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":7168,"tokens_per_second_ceiling":16},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":31744,"tokens_per_second_ceiling":20},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":7168,"tokens_per_second_ceiling":15},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":31744,"tokens_per_second_ceiling":18},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"fits","max_context_tokens":38912,"tokens_per_second_ceiling":17},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":92160,"tokens_per_second_ceiling":22},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":116736,"tokens_per_second_ceiling":25},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":31},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":38},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":46080,"tokens_per_second_ceiling":25},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":70656,"tokens_per_second_ceiling":28},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":100352,"tokens_per_second_ceiling":35},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":124928,"tokens_per_second_ceiling":43},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":46080,"tokens_per_second_ceiling":41},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":70656,"tokens_per_second_ceiling":47},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":100352,"tokens_per_second_ceiling":57},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":124928,"tokens_per_second_ceiling":71},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":7},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":10},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":11},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":14},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"fits","max_context_tokens":38912,"tokens_per_second_ceiling":5},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":92160,"tokens_per_second_ceiling":7},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":116736,"tokens_per_second_ceiling":8},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":9},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":12},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"fits","max_context_tokens":125952,"tokens_per_second_ceiling":3},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":3},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":4},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":5},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":6},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"fits","max_context_tokens":38912,"tokens_per_second_ceiling":2},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":92160,"tokens_per_second_ceiling":3},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":116736,"tokens_per_second_ceiling":4},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":4},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":5},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"fits","max_context_tokens":38912,"tokens_per_second_ceiling":6},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":92160,"tokens_per_second_ceiling":7},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":116736,"tokens_per_second_ceiling":9},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":10},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":13},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":4},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":8},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":10},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":11},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":14},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":150.03,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":6},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":81.01,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":11},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":63.39,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":15},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":55.32,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":17},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":45.77,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":21},
    {"model_id":"deepseek-r1-distill-llama-70b","model":"DeepSeek-R1 Distill Llama 70B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":37.69,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":25},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":477},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":429},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"tight","max_context_tokens":111616,"tokens_per_second_ceiling":394},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":510},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":589},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":721},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":890},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"tight","max_context_tokens":131072,"tokens_per_second_ceiling":580},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":670},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":820},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1012},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"tight","max_context_tokens":131072,"tokens_per_second_ceiling":953},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1101},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1348},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":1663},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":176},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":228},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":263},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":322},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":397},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"tight","max_context_tokens":111616,"tokens_per_second_ceiling":120},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":155},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":179},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":220},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":271},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":60},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":78},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":90},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":110},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":136},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"tight","max_context_tokens":111616,"tokens_per_second_ceiling":56},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":73},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":84},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":103},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":127},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"tight","max_context_tokens":111616,"tokens_per_second_ceiling":135},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":175},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":202},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":247},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":305},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":96},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":180},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":233},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":269},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":329},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":407},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":166.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":140},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":88.64,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":264},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":68.75,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":341},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":59.63,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":394},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":48.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":483},
    {"model_id":"qwen3-coder-next","model":"Qwen3-Coder-Next 80B-A3B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":39.74,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":596},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"tight","max_context_tokens":115712,"tokens_per_second_ceiling":346},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":424},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":523},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"tight","max_context_tokens":131072,"tokens_per_second_ceiling":483},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":595},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"tight","max_context_tokens":131072,"tokens_per_second_ceiling":793},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":978},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"tight","max_context_tokens":131072,"tokens_per_second_ceiling":104},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":134},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":155},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":189},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":234},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"tight","max_context_tokens":115712,"tokens_per_second_ceiling":106},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":129},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":159},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":46},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":53},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":65},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":80},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"tight","max_context_tokens":115712,"tokens_per_second_ceiling":49},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":61},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":75},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"tight","max_context_tokens":115712,"tokens_per_second_ceiling":119},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":145},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":179},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":56},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":106},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":137},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":158},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":194},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":239},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":244.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":82},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":129.79,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":155},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":100.58,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":201},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":87.2,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":232},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":71.38,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":284},
    {"model_id":"gpt-oss-120b","model":"gpt-oss 120B","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":58,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":350},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-3060-12","gpu":"RTX 3060 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4060-ti-16","gpu":"RTX 4060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4070","gpu":"RTX 4070 12 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-3090","gpu":"RTX 3090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-4090","gpu":"RTX 4090 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-5090","gpu":"RTX 5090 32 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-6000-ada","gpu":"RTX 6000 Ada 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"l40s","gpu":"L40S 48 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"tight","max_context_tokens":121856,"tokens_per_second_ceiling":272},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":333},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-pro-6000","gpu":"RTX PRO 6000 Blackwell 96 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":411},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"tight","max_context_tokens":131072,"tokens_per_second_ceiling":379},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"a100-80","gpu":"A100 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":467},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"tight","max_context_tokens":131072,"tokens_per_second_ceiling":622},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"h100-sxm","gpu":"H100 SXM 80 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":768},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"tight","max_context_tokens":131072,"tokens_per_second_ceiling":81},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":105},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":121},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":149},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m2-ultra","gpu":"Mac Studio M2 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":183},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"tight","max_context_tokens":121856,"tokens_per_second_ceiling":83},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":101},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m4-max","gpu":"MacBook Pro M4 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":125},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rtx-5060-ti-16","gpu":"RTX 5060 Ti 16 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"rx-7900-xtx","gpu":"Radeon RX 7900 XTX 24 GB","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":36},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":41},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":51},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"dgx-spark","gpu":"DGX Spark 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":63},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"tight","max_context_tokens":121856,"tokens_per_second_ceiling":39},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":48},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"ryzen-ai-max-395","gpu":"Ryzen AI Max+ 395 128 GB (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":59},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"no","max_context_tokens":0,"tokens_per_second_ceiling":null},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"tight","max_context_tokens":121856,"tokens_per_second_ceiling":93},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":114},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m5-max","gpu":"MacBook Pro M5 Max (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":141},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":44},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":83},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":108},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":124},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":152},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m3-ultra","gpu":"Mac Studio M3 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":188},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"FP16","context_tokens":8192,"need_gb":248.21,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":65},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q8_0","context_tokens":8192,"need_gb":131.87,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":122},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q6_K","context_tokens":8192,"need_gb":102.17,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":158},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q5_K_M","context_tokens":8192,"need_gb":88.56,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":182},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q4_K_M","context_tokens":8192,"need_gb":72.47,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":223},
    {"model_id":"mistral-small-4-119b","model":"Mistral Small 4 119B (MoE)","gpu_id":"m5-ultra","gpu":"Mac Studio M5 Ultra (unified)","quant":"Q3_K_M","context_tokens":8192,"need_gb":58.86,"verdict":"fits","max_context_tokens":131072,"tokens_per_second_ceiling":275}
  ]
}
