Devstral Small 2 (24B)
Mistral AI · Providers EU
Not hosted — mock backend
Open weights
European provider
Sovereignty and hosting
Not hosted — mock backend
Not hosted on LLM EU on this deployment: requests are answered by the built-in mock backend in region dev-mock. No weights run.
Summary
Devstral Small 2 is a 24B Apache-2.0 model fine-tuned for software-engineering agents, built with All Hands AI. The model card reports 68.0% on SWE-bench Verified and 55.7% on SWE-bench Multilingual. It is text-only: the vision encoder of its Mistral Small base was removed during fine-tuning.
Specifications
- Provider
- Mistral AI (FR)
- Context window
- 262k
- Parameters
- 24.01B
- Modalities
- text, code
- License
- apache-2.0 · licence
- Open weights
- yes
- Release date
- 28 Nov 2025
- Tasks
- coding, rag, reasoning
Languages
Strong: en
Adequate: fr, de, es, it, pt, zh, ja, ko
Languages
Strong
- en
Adequate
- fr
- de
- es
- it
- pt
- zh
- ja
- ko
Not verified: the provider has not documented this claim. — a language is listed as strong only where a source claims it; the card links the evidence where one exists.
When to use it
Strengths
- agentic coding: multi-file edits and tool use in repositories
- long 256k-token context window for large codebases
- Apache-2.0 licence permits commercial self-hosting
- vendor-published SWE-bench Multilingual score
Limits
- text-only, no image input
- benchmark scores are self-reported by the vendor
- no evaluation has been run by LLM EU