Magistral Small 1.1
Magistral Small 1.1 updates the behavior and response format of Small 1.0. It retains the 24B Mistral Small 3.1 foundation and Apache 2.0 weights.
A revised reasoning format
Dedicated THINK tokens separate reasoning content from the final response. The reasoning instruction moves into the system prompt. The card also reports better Markdown and LaTeX formatting, shorter answers to easy prompts and fewer endless generations.Magistral Small 1.1 official model card
Similar capability, different behavior
Mistral describes overall benchmark performance as roughly similar to Small 1.0. The revision is useful when output parsing and generation behavior matter. Its weight card still warns about degradation beyond 40K despite describing a 128K architecture window.Magistral Small 1.1 official model card
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | Downloadable |
| Model identifiers | Magistral-Small-2507magistral-small-2507 |
| Context window | 40,000 tokens Historical API advertises 40k; the card recommends 40k despite a 128k architecture window. Magistral Small 1.1 model and lifecycle record |
| Weight license | Apache 2.0Magistral Small 1.1 official model card |
| Access and lifecycle | Hosted API retired on 30 November 2025. Downloadable weights remain listed.Magistral Small 1.1 model and lifecycle recordMagistral Small 1.1 official model card |
| Parameters | 24 billionMagistral Small 1.1 official model card |