Magistral Small at a glance
Magistral Small (24B, Apache 2.0) and the enterprise Magistral Medium reason step by step across many languages, with a fast Think mode in Le Chat.
What changed with Magistral Small
On June 10, 2025, Mistral AI announced Magistral, its first family of reasoning models. Magistral Small is a 24B open-weight model released under Apache 2.0 on Hugging Face, and Magistral Medium is a more powerful enterprise version offered in preview in Le Chat and on La Plateforme, plus Amazon SageMaker.
Mistral reported 73.6% on AIME 2024 for Medium and 70.7% for Small, and emphasized that the chain of thought works natively across languages including English, French, Spanish, German, Italian, Arabic, Russian and Simplified Chinese. With Flash Answers in Le Chat, it said Magistral Medium reaches up to ten times the token throughput of most competitors.
Where to compare Magistral Small
Independent leaderboards and official resources for comparing this model yourself. Rankings change often; each site shows its own current data.
General leaderboards
Why Magistral Small matters for businesses
For businesses, Magistral offered traceable, multilingual reasoning from a European provider, and the Apache-licensed Small version can be self-hosted for tasks such as legal research or planning where you want to inspect how an answer was reached.
Sources for this story
- Magistral (opens in a new tab)Official
Mistral AI · mistral.ai
- mistralai/Magistral-Small-2506 (opens in a new tab)Official
Hugging Face · huggingface.co
- Magistral Medium 1 model page (opens in a new tab)Benchmark
Artificial Analysis · artificialanalysis.ai
Last verified: . IA Compa summarizes public announcements for business readers. We are not affiliated with the companies mentioned. Always confirm details, pricing, and availability with the official source.
- #llm
- #mistral
- #reasoning
- #open-weights
- #multilingual
