Mistral just launched Magistral, its first reasoning AI model family. It tackles complex problems step-by-step like OpenAI’s o3 and Google’s Gemini 2.5 Pro.
Magistral ships in two versions: Magistral Small with 24 billion parameters, downloadable on Hugging Face under an Apache 2.0 license, and Magistral Medium, available in preview on Mistral’s Le Chat chatbot and API, plus partner clouds.
The company says Magistral is built for enterprise needs such as structured calculations, programmatic logic, and decision trees. They claim it offers clear, traceable reasoning in user-friendly language. 
Mistral stated:
“[Magistral is] suited for a wide range of enterprise use cases, from structured calculations and programmatic logic to decision trees and rule-based systems,”
“[The models are] fine-tuned for multi-step logic, improving interpretability and providing a traceable thought process in the user’s language.”
Founded in 2023, Mistral has raised over €1.1 billion ($1.24 billion) from investors like General Catalyst. The lab has lagged in reasoning AI compared to rivals.
Benchmarks show Magistral Medium falls short of Gemini 2.5 Pro and Anthropic’s Claude Opus 4 on physics, math, and science tests like GPQA Diamond and AIME. It also doesn’t beat Gemini 2.5 Pro on the programming benchmark LiveCodeBench.
Speed is Magistral’s main edge. Mistral claims Magistral runs queries 10x faster than competitors on Le Chat and supports many languages—Italian, Arabic, Russian, Simplified Chinese, and more.
Mistral explained:
“Building on our flagship models, Magistral is designed for research, strategic planning, operational optimization, and data-driven decision making,”
“whether executing risk assessment and modelling with multiple factors, or calculating optimal delivery windows under constraints.”
Magistral follows several recent releases: the Mistral Code “vibe coding” client, new coding-focused models, and Le Chat Enterprise, a business chatbot with AI agent tools and integrations for Gmail and SharePoint.
