Mistral Launches Magistral: Its First AI Reasoning Models

French AI lab Mistral has officially entered the AI reasoning space with the launch of its new model family, Magistral. These models are designed to solve complex problems using step-by-step logic, similar to OpenAI’s o3 and Google’s Gemini 2.5 Pro.

Two Versions Available

Mistral released two versions of Magistral:

1. Magistral Small – A lightweight model with 24 billion parameters, available for free under the open-source Apache 2.0 license on Hugging Face.
2. Magistral Medium – A more advanced version, available for preview via Mistral’s Le Chat platform, API, and third-party cloud services.

These models are designed for enterprise use cases like structured calculations, programmatic logic, decision trees, and rule-based systems.

Key Features

Mistral claims that Magistral is:

1. Fine-tuned for multi-step reasoning, offering better interpretability.
2. Capable of delivering traceable thought processes in natural language.
3. Suited for applications such as risk assessment, strategic planning, and data-driven decision making.

Magistral also supports multiple languages including Italian, Arabic, Russian, and Simplified Chinese, and reportedly delivers results at 10x the speed of competitors on the Le Chat platform.

Performance & Limitations

Despite its speed and multilingual support, Magistral Medium underperforms on several AI benchmarks:

1. On GPQA Diamond and AIME (tests for math and science), it lags behind Gemini 2.5 Pro and Claude Opus 4.
2. It also trails Gemini 2.5 Pro on LiveCodeBench, a programming performance test.

Company Background

Founded in 2023, Mistral has quickly risen in the AI industry, raising over €1.1 billion (~$1.24 billion) from investors like General Catalyst. The company has developed several AI products including:

1. Le Chat – A general-purpose chatbot
2. Mistral Code – A client for “vibe coding”
3. Le Chat Enterprise – A business-focused version with tools like an AI agent builder and integrations for Gmail and SharePoint

Conclusion

While Magistral may not yet compete with top-tier reasoning models in terms of raw benchmark performance, Mistral is emphasizing its speed, language coverage, and enterprise usability as key differentiators.