Devstral Small 2 24B
Devstral Small 2 24B needs 16.3 GB of VRAM at Q4_K_M with 8k context, computed from its config.json as checked on 3 November 2026. The smallest card that holds it has 24 GB, and the fastest desktop card that runs it, the RTX 5090, generates about 90 tokens per second.
- Parameters
- 24B
- Context
- 262,144 tokens
- Licence
- Apache 2.0
- Released
- December 2025
- Config checked
- 3 November 2026
Open Devstral Small 2 24B in the LLM checker
Which card size runs it
16.3 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 3 November 2026Which GPU to pick
The smallest card here that loads it at Q4_K_M with 8k of context is the RX 7900 XT (20 GB): 16.3 GB needed, an estimated 40 tokens per second. For room to work, meaning 32,768 tokens of context while staying under 85% of memory, step up to the RTX 5090 Laptop GPU (24 GB), which needs 20.3 GB for that and generates at about 45 tokens per second.
At Q8_0, which is near-lossless, the card is the RTX 5090 (32 GB), needing 28.3 GB. On Apple silicon, a Mac mini M4 Pro holds it with 32,768 tokens of context at an estimated 14 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 20.3 GB, at an estimated 13 tokens per second.
Speed on each card
| GPU | Fits | Est. tokens/s | Feels like | Memory | Bandwidth |
|---|---|---|---|---|---|
| H100 SXM | Yes | ≤ 168 | faster than you can read | 80 GB | 3,350 GB/s |
| A100 | Yes | ≤ 103 | faster than you can read | 80 GB | 2,039 GB/s |
| RTX 5090 | Yes | 90 | faster than you can read | 32 GB | 1,792 GB/s |
| RTX PRO 6000 Blackwell | Yes | 90 | faster than you can read | 96 GB | 1,792 GB/s |
| RTX PRO 5000 Blackwell | Yes | 68 | faster than you can read | 48 GB | 1,344 GB/s |
| RTX 4090 | Yes | 51 | comfortable for chat | 24 GB | 1,008 GB/s |
| RX 7900 XTX | Yes | 48 | comfortable for chat | 24 GB | 960 GB/s |
| RTX 6000 Ada | Yes | 48 | comfortable for chat | 48 GB | 960 GB/s |
| RTX 3090 | Yes | 47 | comfortable for chat | 24 GB | 936 GB/s |
| RTX 5090 Laptop GPU | Yes | 45 | comfortable for chat | 24 GB | 896 GB/s |
| RTX PRO 4500 Blackwell | Yes | 45 | comfortable for chat | 32 GB | 896 GB/s |
| L40S | Yes | 43 | comfortable for chat | 48 GB | 864 GB/s |
| RX 7900 XT | Yes | 40 | comfortable for chat | 20 GB | 800 GB/s |
| RTX A6000 | Yes | 39 | comfortable for chat | 48 GB | 768 GB/s |
| RTX PRO 4000 Blackwell | Yes | 34 | comfortable for chat | 24 GB | 672 GB/s |
| Radeon AI PRO R9700 | Yes | 32 | comfortable for chat | 32 GB | 640 GB/s |
| Arc Pro B70 | Yes | 31 | comfortable for chat | 32 GB | 608 GB/s |
| Arc Pro B60 | Yes | 23 | usable, a little slow | 24 GB | 456 GB/s |
| RTX PRO 4000 Blackwell SFF | Yes | 22 | usable, a little slow | 24 GB | 432 GB/s |
| RTX Spark | Yes | 15 | usable, a little slow | 128 GB (102 for the GPU) | 300 GB/s |
| DGX Spark | Yes | 14 | usable, a little slow | 128 GB (126 for the GPU) | 273 GB/s |
| Ryzen AI Max+ 395 | Yes | 13 | usable, a little slow | 128 GB (96 for the GPU) | 256 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: Devstral Small 2 24B reads only 13.9 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 2 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 3 November 2026Memory by quantisation and context
| Quant | 4,096 ctx | 8,192 ctx | 32,768 ctx |
|---|---|---|---|
| Q8_0 | 27.6 GB · fits 32 GB | 28.3 GB · fits 32 GB | 32.3 GB · fits 48 GB |
| Q5_K_M | 18.9 GB · fits 24 GB | 19.6 GB · fits 24 GB | 23.6 GB · fits 32 GB |
| Q4_K_M | 15.6 GB · fits 24 GB | 16.3 GB · fits 24 GB | 20.3 GB · fits 24 GB |
Weights at this quant: 13.9 GB at Q4. Every extra 1,000 tokens of context adds about 0.164 GB of KV cache at FP16. Try other settings in the VRAM calculator.
About the model
Devstral Small 2 24B has 24B parameters, 40 layers and 8 KV heads of dimension 128. Mistral's agentic coding model, December 2025. It is built to explore a codebase with tools, edit several files and drive a software-engineering agent, and Mistral released its Vibe CLI alongside it.
Text and images 262,144-token window Apache 2.0 December 2025
Reasoning. Answers directly.
Mistral's coding model. The parameter count includes the vision encoder; a text-only GGUF is slightly smaller.
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: 262,144 tokens. Licence: Apache 2.0, a permissive licence. Weights published in December 2025.
Good at
- A coding agent on a single 24 GB card, which is the use Mistral names on the card.
- Reading screenshots and diagrams as part of a coding task.
- Apache 2.0, for a coding model you can ship inside a product.
Watch out for
- It is a specialist. For general conversation a general model of the same size is the better choice.
- Every layer keeps a full-length cache, so long context costs several times what a 2026 hybrid model pays for the same conversation. Repository-scale context is exactly where that bites.
Similar models
The models that need about the same memory as Devstral Small 2 24B 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 |
|---|---|---|---|---|---|
| Devstral Small 2 24B | 24B | 16.3 GB | 0.164 GB | 256k | 2025-12 |
| Mistral Small 3.1 24B | 24B | 16.3 GB | 0.164 GB | 128k | 2025-03 |
| Gemma 4 26B-A4B (MoE) | 25.8B · 3.8B active | 16.4 GB | 0.02 GB | 256k | 2026-03 |
| Qwen3.8 27B | 27.8B | 18 GB | 0.066 GB | 256k | 2026-08 |
| Qwen3.6 27B | 27.8B | 18 GB | 0.066 GB | 256k | 2026-04 |
See every model compared, which models fit on 8 to 96 GB, or the cheapest hour of every cloud card.
Questions
- Can I run Devstral Small 2 24B on a 24 GB card?
- Yes, at Q4_K_M and 8k context it needs about 16.3 GB, which fits a 24 GB card with room.
- How much VRAM does Devstral Small 2 24B need at Q8?
- About 28.3 GB at 8k context, or 32.3 GB at 32k. Q8 is near-lossless; use it when it fits.
- What is the context length of Devstral Small 2 24B?
- 262,144 tokens natively. Holding all of it at Q4_K_M takes about 57.9 GB, of which 42.9 GB is context.
- How fast is Devstral Small 2 24B on an RTX 4090?
- At most about 51 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 mistralai/Devstral-Small-2-24B-Instruct-2512 on Hugging Face, checked 3 November 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.