Qwen2.5 Coder 32B
Qwen2.5 Coder 32B needs 22.4 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 24 GB, a tight fit, and the fastest desktop card that runs it, the RTX 5090, generates about 66 tokens per second.
- Parameters
- 32.8B
- Context
- 32,768 tokens
- Licence
- Apache 2.0
- Released
- November 2024
- Config checked
- 16 October 2026
Open Qwen2.5 Coder 32B in the LLM checker
Which card size runs it
22.4 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 RTX 3090 (24 GB): 22.4 GB needed, an estimated 34 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 A6000 (48 GB), which needs 28.9 GB for that and generates at about 28 tokens per second.
At Q8_0, which is near-lossless, the card is the RTX A6000 (48 GB), needing 38.8 GB. On Apple silicon, a Mac mini M4 Pro holds it with 32,768 tokens of context at an estimated 10 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 28.9 GB, at an estimated 9 tokens per second.
Speed on each card
| GPU | Fits | Est. tokens/s | Feels like | Memory | Bandwidth |
|---|---|---|---|---|---|
| H100 SXM | Yes | ≤ 123 | faster than you can read | 80 GB | 3,350 GB/s |
| A100 | Yes | 75 | faster than you can read | 80 GB | 2,039 GB/s |
| RTX 5090 | Yes | 66 | faster than you can read | 32 GB | 1,792 GB/s |
| RTX PRO 6000 Blackwell | Yes | 66 | faster than you can read | 96 GB | 1,792 GB/s |
| RTX PRO 5000 Blackwell | Yes | 49 | comfortable for chat | 48 GB | 1,344 GB/s |
| RTX 4090 | Tight | 37 | comfortable for chat | 24 GB | 1,008 GB/s |
| RX 7900 XTX | Tight | 35 | comfortable for chat | 24 GB | 960 GB/s |
| RTX 6000 Ada | Yes | 35 | comfortable for chat | 48 GB | 960 GB/s |
| RTX 3090 | Tight | 34 | comfortable for chat | 24 GB | 936 GB/s |
| RTX PRO 4500 Blackwell | Yes | 33 | comfortable for chat | 32 GB | 896 GB/s |
| L40S | Yes | 32 | comfortable for chat | 48 GB | 864 GB/s |
| RTX A6000 | Yes | 28 | comfortable for chat | 48 GB | 768 GB/s |
| RTX PRO 4000 Blackwell | Tight | 25 | usable, a little slow | 24 GB | 672 GB/s |
| Radeon AI PRO R9700 | Yes | 24 | usable, a little slow | 32 GB | 640 GB/s |
| Arc Pro B70 | Yes | 22 | usable, a little slow | 32 GB | 608 GB/s |
| Arc Pro B60 | Tight | 17 | usable, a little slow | 24 GB | 456 GB/s |
| RTX PRO 4000 Blackwell SFF | Tight | 16 | usable, a little slow | 24 GB | 432 GB/s |
| RTX Spark | Yes | 11 | slow; fine for batch jobs | 128 GB (102 for the GPU) | 300 GB/s |
| DGX Spark | Yes | 10 | slow; fine for batch jobs | 128 GB (126 for the GPU) | 273 GB/s |
| Ryzen AI Max+ 395 | Yes | 9 | slow; fine for batch jobs | 128 GB (96 for the GPU) | 256 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: Qwen2.5 Coder 32B reads only 19 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 figure marked ≤ is 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 | 37.7 GB · fits 48 GB | 38.8 GB · fits 48 GB | 45.2 GB · fits 48 GB |
| Q5_K_M | 25.8 GB · fits 32 GB | 26.9 GB · fits 32 GB | 33.3 GB · fits 48 GB |
| Q4_K_M | 21.4 GB · fits 24 GB | 22.4 GB · fits 24 GB | 28.9 GB · fits 32 GB |
Qwen3-Coder-Next 80B-A3B (MoE) is the newer model in this family. It needs about 48.9 GB at the same setting.
Weights at this quant: 19 GB at Q4. Every extra 1,000 tokens of context adds about 0.262 GB of KV cache at FP16. Try other settings in the VRAM calculator.
About the model
Qwen2.5 Coder 32B has 32.8B parameters, 64 layers and 8 KV heads of dimension 128. Alibaba's coding model from November 2024, and the open coding model of choice through most of 2025.
Text only 32,768-token window Apache 2.0 November 2024
Reasoning. Answers directly; there is no reasoning mode.
This is a coding model, tuned for writing and editing code and for agentic tools that read whole repositories, which is where a long context earns its memory.
Native context window: 32,768 tokens. Licence: Apache 2.0, a permissive licence. Weights published in November 2024.
Good at
- Code completion and single-file work on a 24 GB card.
- Editor integrations that were built and tuned against it.
Watch out for
- A 32,768-token native window, which is small for repository-scale work.
- It predates agentic coding as a training target. Devstral Small 2 on the same card, or Qwen3-Coder-Next on a larger one, were built for it.
- Every layer keeps a full-length cache, so long context costs several times what a 2026 hybrid model pays for the same conversation.
Similar models
The models that need about the same memory as Qwen2.5 Coder 32B 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 |
|---|---|---|---|---|---|
| Qwen2.5 Coder 32B | 32.8B | 22.4 GB | 0.262 GB | 32k | 2024-11 |
| Qwen3 32B | 32.8B | 22.4 GB | 0.262 GB | 40k | 2025-04 |
| DeepSeek-R1 Distill Qwen 32B | 32.8B | 22.4 GB | 0.262 GB | 128k | 2025-01 |
| Qwen3.6 35B-A3B (MoE) | 36B · 3B active | 22.4 GB | 0.02 GB | 256k | 2026-04 |
| Gemma 4 31B | 31.3B | 20.9 GB | 0.082 GB | 256k | 2026-03 |
See every model compared, which models fit on 8 to 96 GB, or the cheapest hour of every cloud card.
Questions
- Can I run Qwen2.5 Coder 32B on a 24 GB card?
- Yes, at Q4_K_M and 8k context it needs about 22.4 GB, which fits a 24 GB card tightly.
- How much VRAM does Qwen2.5 Coder 32B need at Q8?
- About 38.8 GB at 8k context, or 45.2 GB at 32k. Q8 is near-lossless; use it when it fits.
- What is the context length of Qwen2.5 Coder 32B?
- 32,768 tokens natively. That is short by current standards, and it is a limit of the model rather than of your GPU: spare memory does not extend it. Qwen documents YaRN scaling to stretch it further, at some cost in quality on short prompts.
- Is there a newer model than Qwen2.5 Coder 32B?
- Yes: Qwen3-Coder-Next 80B-A3B (MoE), published in January 2026. At the same Q4_K_M and 8k setting it needs about 48.9 GB against 22.4 GB for this one. It is a much larger model, so it is a successor in name rather than a drop-in replacement.
- How fast is Qwen2.5 Coder 32B on an RTX 4090?
- At most about 37 tokens per second at Q4, single stream, which is comfortable for chat. That is a ceiling worked out from the card's memory bandwidth, not a measurement.
- Architecture values from config.json in Qwen/Qwen2.5-Coder-32B-Instruct 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.