gpt-oss 120B
gpt-oss 120B needs 71.4 GB of VRAM at Q4_K_M with 8k context, computed from its config.json as checked on 16 October 2026. The smallest card that holds it has 80 GB, a tight fit, and the fastest desktop card that runs it, the RTX PRO 6000 Blackwell, generates at most 424 tokens per second.
- Parameters
- 117B, 5.1B active
- Context
- 131,072 tokens
- Licence
- Apache 2.0
- Released
- August 2025
- Config checked
- 16 October 2026
Open gpt-oss 120B in the LLM checker
Which card size runs it
71.4 GB against each memory size. A model may use 95% of a card, and only what is left over, or missing, is coloured; tight means above 85%, with little room for a longer conversation.
Other quantisations, context lengths or a Q8 KV cache: the VRAM calculator.
Computed 16 October 2026Which GPU to pick
The smallest card here that loads it at Q4_K_M with 8k of context is the A100 (80 GB): 71.4 GB needed, an estimated 483 tokens per second, and a tight fit that leaves little for the conversation. For room to work, meaning 32,768 tokens of context while staying under 85% of memory, step up to the RTX PRO 6000 Blackwell (96 GB), which needs 72.3 GB for that and generates at about 424 tokens per second.
At Q8_0 it does not fit any single card listed here. On Apple silicon, a MacBook Pro M4 Max holds it with 32,768 tokens of context at an estimated 129 tokens per second; prompt processing is slower than on Nvidia, which long documents make obvious. Among unified-memory desktops, the Ryzen AI Max+ 395 holds it at 32,768 tokens in 72.3 GB, at an estimated 61 tokens per second.
Speed on each card
| GPU | Fits | Est. tokens/s | Feels like | Memory | Bandwidth |
|---|---|---|---|---|---|
| H100 SXM | Tight | ≤ 793 | faster than you can read | 80 GB | 3,350 GB/s |
| A100 | Tight | ≤ 483 | faster than you can read | 80 GB | 2,039 GB/s |
| RTX PRO 6000 Blackwell | Yes | ≤ 424 | faster than you can read | 96 GB | 1,792 GB/s |
| RTX Spark | Yes | 71 | faster than you can read | 128 GB (102 for the GPU) | 300 GB/s |
| DGX Spark | Yes | 65 | faster than you can read | 128 GB (126 for the GPU) | 273 GB/s |
| Ryzen AI Max+ 395 | Yes | 61 | faster than you can read | 128 GB (96 for the GPU) | 256 GB/s |
| RTX 6000 Ada | No | - | does not fit | 48 GB | 960 GB/s |
| L40S | No | - | does not fit | 48 GB | 864 GB/s |
| RTX A6000 | No | - | does not fit | 48 GB | 768 GB/s |
| RTX PRO 5000 Blackwell | No | - | does not fit | 48 GB | 1,344 GB/s |
| RTX 5090 | No | - | does not fit | 32 GB | 1,792 GB/s |
| RTX PRO 4500 Blackwell | No | - | does not fit | 32 GB | 896 GB/s |
| Radeon AI PRO R9700 | No | - | does not fit | 32 GB | 640 GB/s |
| Arc Pro B70 | No | - | does not fit | 32 GB | 608 GB/s |
| RTX 3090 | No | - | does not fit | 24 GB | 936 GB/s |
| RTX 4090 | No | - | does not fit | 24 GB | 1,008 GB/s |
| RX 7900 XTX | No | - | does not fit | 24 GB | 960 GB/s |
| RTX PRO 4000 Blackwell | No | - | does not fit | 24 GB | 672 GB/s |
| RTX PRO 4000 Blackwell SFF | No | - | does not fit | 24 GB | 432 GB/s |
| Arc Pro B60 | No | - | does not fit | 24 GB | 456 GB/s |
| RX 7900 XT | No | - | does not fit | 20 GB | 800 GB/s |
| RTX 4060 Ti | No | - | does not fit | 16 GB | 288 GB/s |
| RTX 5060 Ti | No | - | does not fit | 16 GB | 448 GB/s |
| RTX 5080 | No | - | does not fit | 16 GB | 960 GB/s |
| RTX 5070 Ti | No | - | does not fit | 16 GB | 896 GB/s |
| RTX 4080 SUPER | No | - | does not fit | 16 GB | 736 GB/s |
| RX 9070 XT | No | - | does not fit | 16 GB | 640 GB/s |
| RX 9060 XT | No | - | does not fit | 16 GB | 320 GB/s |
| RTX 3060 | No | - | does not fit | 12 GB | 360 GB/s |
| RTX 4070 | No | - | does not fit | 12 GB | 504 GB/s |
| RTX 5070 | No | - | does not fit | 12 GB | 672 GB/s |
| Arc B580 | No | - | does not fit | 12 GB | 456 GB/s |
Single-stream decode, from memory bandwidth at 70% efficiency. Prompt processing and batching not included. These are ceilings, not measurements: gpt-oss 120B reads only 3 GB of weights per token at Q4, so little that the per-token costs the formula leaves out decide much of the real speed. The 3 figures marked ≤ are where a real runtime falls furthest below the number shown. See the speed estimator for other quantisations and Apple hardware, or every card compared if you are choosing hardware rather than a model.
Computed 16 October 2026Memory by quantisation and context
| Quant | 4,096 ctx | 8,192 ctx | 32,768 ctx |
|---|---|---|---|
| Q8_0 | 129.6 GB · fits multi-GPU | 129.8 GB · fits multi-GPU | 130.7 GB · fits multi-GPU |
| Q5_K_M | 87 GB · fits 96 GB | 87.2 GB · fits 96 GB | 88.1 GB · fits 96 GB |
| Q4_K_M | 71.2 GB · fits 80 GB | 71.4 GB · fits 80 GB | 72.3 GB · fits 80 GB |
Weights at this quant: 67.9 GB at Q4. Every extra 1,000 tokens of context adds about 0.037 GB of KV cache at FP16, on top of a fixed 0.005 GB that the sliding-window layers hold once their window is full. Try other settings in the VRAM calculator.
About the model
gpt-oss 120B has 117B parameters with 5.1B active per token (mixture of experts). It has 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. The larger of OpenAI's open-weight models, August 2025: 117B of weights, 5.1B read per token, shipped in MXFP4 so that it fits a single 80 GB card.
Text only 131,072-token window Apache 2.0 August 2025 5.1B of 117B active
Reasoning. Always reasons, at low, medium or high effort, set in the system prompt.
Ships in MXFP4; OpenAI states it runs on a single 80 GB GPU. Choose the 4-bit row.
Native context window: 131,072 tokens. Licence: Apache 2.0, a permissive licence. Weights published in August 2025.
Good at
- One 80 GB card, which is what OpenAI designed it for.
- A 128 GB Mac, where mixture-of-experts models make the most of slow, plentiful memory.
- Tool use and agent work under Apache 2.0.
Watch out for
- It expects the harmony chat format.
- MXFP4 is its native precision; there is no higher-quality version to step up to.
- Text only.
Similar models
The models that need about the same memory as gpt-oss 120B at Q4_K_M and 8k, which makes them the real alternatives on whatever card you have. The context column is what each extra 1,000 tokens costs, and it is where models of the same size differ most.
| Model | Size | Needs | Per 1k context | Window | Released |
|---|---|---|---|---|---|
| gpt-oss 120B | 117B · 5.1B active | 71.4 GB | 0.037 GB | 128k | 2025-08 |
| Mistral Small 4 119B (MoE) | 119B · 6.5B active | 72.5 GB | 0.023 GB | 256k | 2026-03 |
| Qwen3-Coder-Next 80B-A3B (MoE) | 79.7B · 3B active | 48.9 GB | 0.025 GB | 256k | 2026-01 |
| Nemotron 3.5 Bolt 74B-A9B (MoE) | 74.9B · 9B active | 45.9 GB | 0.008 GB | 256k | 2026-10 |
| DeepSeek-R1 Distill Llama 70B | 70.6B | 45.8 GB | 0.328 GB | 128k | 2025-01 |
See every model compared, which models fit on 8 to 96 GB, or the cheapest hour of every cloud card.
Questions
- Can I run gpt-oss 120B on a 24 GB card?
- Not comfortably. At Q4_K_M and 8k context it needs about 71.4 GB. The smallest tier that fits is 80 GB.
- How much VRAM does gpt-oss 120B need at Q8?
- About 129.8 GB at 8k context, or 130.7 GB at 32k. Q8 is near-lossless; use it when it fits.
- What is the context length of gpt-oss 120B?
- 131,072 tokens natively. Holding all of it at Q4_K_M takes about 75.9 GB, of which 4.8 GB is context.
- Why does gpt-oss 120B need so little memory for long context?
- Because most of its layers do not keep a cache that grows. It has 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. At 32k tokens the context costs about 1.2 GB; if all 36 layers kept a conventional full-length cache, the same conversation would cost about 2.4 GB. That saving depends on the runtime implementing the sliding-window cache. llama.cpp and vLLM do; a build that does not will use far more.
- How fast is gpt-oss 120B on an RTX 4090?
- It does not fit a 4090 at Q4 and 8k context, so speed would collapse to CPU offloading. Use a larger card or a smaller quantisation.
- Architecture values from config.json in openai/gpt-oss-120b on Hugging Face, checked 16 October 2026. Every figure on this page is computed from those values; check them before buying hardware for this model.
- Estimates assume a single conversation on a card that is otherwise free. A desktop on the same GPU takes 0.5 to 2 GB.