Devstral Small
Mistral AI · Code · 24B · 131k context · Released 25 July 2025
Devstral Small is fine-tuned from Mistral Small 3.1 for agentic coding: exploring codebases, editing multiple files, and using tools, rather than one-shot code generation. Version 1.1 is the current release. It is a good fit for people running coding agents locally who want a permissive licence.
Strengths
- Purpose-built for coding agents and multi-file, tool-driven work
- Runs on a single 24GB card or a 32GB Mac
- Apache 2.0, so usable commercially without conditions
Weaknesses
- Specialised for agentic workflows, so less suited to plain chat
- A 24B model still trails the largest coding models on the hardest tasks
- Gets the most out of an agent harness like OpenHands, which adds setup
Hardware requirements
| Quantisation | Approx. VRAM | Notes |
|---|---|---|
| Q4_K_M | ~14GB | Fits a 16GB card with short context |
| Q5_K_M | ~17GB | A good balance on a 24GB card |
| Q8_0 | ~25GB | Near-lossless, needs 32GB or more |
| FP16 | ~48GB | Full precision, server or multi-GPU territory |
Also runs on CPU (slower). Optimised builds available for Apple Silicon.
What you'd need to run this
Roughly what a machine to run this would need, at up to three levels of quality. Memory is the deciding factor.
Minimum to run it
Q4_K_M · ~14GB needed
One 24GB GPU
NVIDIA Tesla P40or a Mac or mini-PC with unified memory, if you prefer no discrete GPU, Mac mini M5 Pro .
32–64GB of system RAM alongside the card.
around £700–£1,100
What else 24GB runs →For good quality
Q5_K_M · ~17GB needed
One 24GB GPU
NVIDIA Tesla P40or a Mac or mini-PC with unified memory, if you prefer no discrete GPU, Mac mini M5 Pro .
32–64GB of system RAM alongside the card.
around £700–£1,100
What else 24GB runs →Best quality
FP16 · ~48GB needed
64GB of unified memory
Mac mini M4 Proor an 80GB-class data-centre card, which is usually rented by the hour, NVIDIA A100 80GB .
Unified memory is shared with the model, so it is already counted above.
Licence
Apache 2.0 — read the licence
Benchmarks
| Benchmark | Score | Source | As of |
|---|---|---|---|
| SWE-bench Verified | 53.6% | Mistral AI model card (Devstral Small 1.1) | July 2025 |
How it compares
How this model’s reported scores sit against other models we cover, on the same benchmarks. This model is highlighted.
SWE-bench Verified
higher is better- Muse Glimmer 30B 76.0%
Meta model card · August 2026
- gpt-oss-120b 62.4%
OpenAI gpt-oss model card (high reasoning) · August 2025
- gpt-oss-20b 60.7%
OpenAI gpt-oss model card (high reasoning) · August 2025
- Devstral Small 53.6%
Mistral AI model card (Devstral Small 1.1) · July 2025
- Qwen3-Coder 30B-A3B 51.6
Qwen (official repo, OpenHands scaffold, 100 turns) · August 2025
- Nemotron 3.5 Lightning 51.56
NVIDIA (BF16) · August 2026
Devstral Small: common questions
- What hardware do I need to run Devstral Small?
- At its most compressed (Q4_K_M) it needs roughly 14GB of VRAM, and about 17GB for good quality. VRAM figures are approximate and depend on context length and settings.
- Is Devstral Small free for commercial use?
- Yes. Devstral Small is licensed under Apache 2.0, which permits commercial use with no meaningful conditions.
- Can I run Devstral Small on Apple Silicon?
- Yes. Devstral Small has builds optimised for Apple Silicon, through MLX or GGUF on a Mac.
- Does Devstral Small run on CPU?
- Yes, Devstral Small can run on the CPU, though generation is slower than on a GPU.
- What is Devstral Small's context window?
- Devstral Small has a context window of 131,072 tokens, about 131k.
Availability
- Official page
- Hugging Face
- ollama run devstral:24b
Where to get quantised weights
Some of the best quantised weights are made by the community, not the model’s authors. Look this model up on these providers:
- Unsloth GGUF (Dynamic 2.0, imatrix)
Dynamic and imatrix GGUF quants that often hold quality better than a plain quant at the same bit-width, especially at 4-bit and below.
- Bartowski GGUF Q2-Q8 (imatrix)
A wide, reliable range of imatrix GGUF quants, typically Q2 through Q8.
- MLX community MLX 4-bit and 8-bit
MLX quants for Apple Silicon, usually 4-bit and 8-bit.
Recommended for
- Running local coding agents on a single 24GB card
- OpenHands and similar agent harnesses
- Commercial coding work needing a permissive licence
Related guides
Glossary
Catalogue entry last verified 30 July 2026. Specifications change; verify anything you are about to spend money on.