Mistral’s First Reasoning Model, Magistral, Launches with Large and Small Apache 2.0 Version
- French AI startup Mistral launched Magistral, its first reasoning model family, on Tuesday with dual releases both open-source and proprietary.
- Mistral developed Magistral to address complex problem solving with step-by-step reasoning and traceability, amidst a competitive landscape including OpenAI and DeepSeek.
- Magistral Small is a 24-billion parameter open-source model under Apache 2.0, while Magistral Medium offers enhanced enterprise features and is cloud-available on platforms like Amazon SageMaker.
- Magistral Medium scores 73.6% on the AIME-24 math benchmark, improving to 90% with majority voting, and offers lower output costs than OpenAI and Gemini 2.5 Pro at $5 per million tokens.
- Mistral signals an open yet competitive AI reasoning future by supporting multilingual, traceable reasoning aimed at law, finance, and healthcare, positioning Magistral for high-precision professional use cases.
Insights by Ground AI
Does this summary seem wrong?
32 Articles
32 Articles
All
Left
2
Center
5
Right
2


Mistral releases a pair of AI reasoning models
Mistral released Magistral, its first family of reasoning models. Like other reasoning models — e.g. OpenAI's o3 and Google's Gemini 2.5 Pro — Magistral works through problems step-by-step for improved consistency and reliability across topics such as math and physics.
·United States
Read Full ArticleFrench AI startup Mistral to launch AI reasoning model that thinks in multiple languages - Tech Startups
French AI startup Mistral on Tuesday announced it’s launching its first reasoning model to compete take rivals like OpenAI and China’s DeepSeek. Unlike the other reasoning models, Mistral’s new model will bring something fresh to the table that its competitors […] The post French AI startup Mistral to launch AI reasoning model that thinks in multiple languages first appeared on Tech Startups.
Coverage Details
Total News Sources32
Leaning Left2Leaning Right2Center5Last UpdatedBias Distribution56% Center
Bias Distribution
- 56% of the sources are Center
56% Center
L 22%
C 56%
R 22%
Factuality
To view factuality data please Upgrade to Premium