🚨 Breaking News 🚨
Mistral AI just unveiled Magistral!
It’s their first reasoning language models, designed for transparent, domain-specific, and multilingual reasoning.
The release includes both an open-source and an enterprise variant, aiming to set new standards in step-by-step, verifiable AI reasoning.
Key Features
1. Dual Model Release
Magistral Small: 24B parameter open-source model, available under Apache 2.0 license.
Magistral Medium: More powerful, enterprise-focused version, accessible via API, Le Chat, and cloud platforms.
2. Reasoning Performance
Magistral Medium: 73.6% on AIME2024, 90% with majority voting @64.
Magistral Small: 70.7% and 83.3% respectively.
3. Transparent, Traceable Reasoning
Fine-tuned for multi-step logic and interpretability.…
🚨 Breaking News 🚨
Mistral AI just unveiled Magistral!
It’s their first reasoning language models, designed for transparent, domain-specific, and multilingual reasoning.
The release includes both an open-source and an enterprise variant, aiming to set new standards in step-by-step, verifiable AI reasoning.
Key Features
1. Dual Model Release
Magistral Small: 24B parameter open-source model, available under Apache 2.0 license.
Magistral Medium: More powerful, enterprise-focused version, accessible via API, Le Chat, and cloud platforms.
2. Reasoning Performance
Magistral Medium: 73.6% on AIME2024, 90% with majority voting @64.
Magistral Small: 70.7% and 83.3% respectively.
3. Transparent, Traceable Reasoning
Fine-tuned for multi-step logic and interpretability.
Delivers a clear, auditable thought process in the user’s language.
4. Multilingual Dexterity
High-fidelity reasoning in English, French, Spanish, German, Italian, Arabic, Russian, and Simplified Chinese.
5. Enterprise-Ready
Suited for structured calculations, programmatic logic, decision trees, and rule-based systems.
Supports legal, finance, healthcare, and government use cases with compliance-friendly, traceable outputs.
6. 10x Faster Responses
“Flash Answers” in Le Chat provide up to 10x faster token throughput than most competitors.
7. Versatile Applications
Legal research, financial forecasting, software development, creative writing, and more.
8. Open-Source Community Focus
Magistral Small is open for self-deployment and community-driven innovation.
Early open models have already powered projects like ether0 and DeepHermes 3.
Multilingual Dexterity
High-fidelity reasoning in English, French, Spanish, German, Italian, Arabic, Russian, and Simplified Chinese.
10x Faster Reasoning With Le Chat
“Flash Answers” in Le Chat provide up to 10x faster token throughput than most competitors.