Open-weight agentic coding model built on Mistral Small 3.1, which Mistral said set a new state of the art for open models on SWE-bench Verified at release.

Benchmarks

What its maker published, and what anyone else measured. Every figure links the document it came from.

benchmark	published	measured
SWE-bench	53.6	—

Lineage

No recorded lineage — root or standalone model.

view in full graph →

Metadata

org	Mistral AI
released	2025-07-10
params	24B
license	apache-2.0
hf id	mistralai/Devstral-Small-2507