AI BriefingMistral AI Unveils Its First Reasoning Model 'Magistral'
Key point
Mistral AI has unveiled Magistral, its first reasoning model with strengths in domain-specific and multilingual reasoning.
Details
Mistral AI has unveiled Magistral, its first reasoning model equipped with domain-specific knowledge, transparent reasoning processes, and outstanding multilingual capabilities. This model is optimized to analyze and solve complex problems step by step.
The model is available in two versions: Magistral Small, an open-source version with 24B parameters, and Magistral Medium, an enterprise version with more powerful performance.
Looking at the performance metrics, Magistral Medium recorded 73.6% on AIME2024, and showed a high accuracy of 90% when majority voting (majority voting @64) was applied. Magistral Small also achieved 70.7% on AIME2024, and 83.3% with majority voting.
Key features are as follows:
- Transparent Reasoning: Unlike typical models, it provides a traceable chain of thought (Chain-of-Thought) in the user's language, enhancing interpretability.
- Multilingual Flexibility: It maintains advanced reasoning performance across various languages including English, French, Spanish, German, Italian, Arabic, Russian, and Chinese.
- Overwhelming Speed: Through Le Chat's Flash Answers feature, it delivers up to 10x faster token throughput compared to competitors, supporting real-time reasoning.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.