Mistral, the ambitious AI research lab from France, has thrown its hat into the reasoning model ring.
On Tuesday, Mistral introduced AI models tailor made for logical problem solving. These models work through complex topics like math and science with a step at a time approach, providing more consistent results and better transparency in how answers are reached.
Two versions make up the Magistral lineup. Magistral Small relies on 24 billion parameters and is available for direct download through Hugging Face, accompanied by an open source license that allows developers to experiment freely. In contrast, Magistral Medium is accessible in preview form via the Mistral Le Chat chatbot platform, through the company’s API, and on various cloud partners.
Mistral says Magistral shines in areas that require structured calculation and logical flow, making it an option for enterprises that depend on rule systems, decision processes, and multiple sequential steps. The company claims these models have tuned multi step thinking that is easier for users to follow, with clear explanations available in many languages.
Mistral Looks to Catch Up in Reasoning AI
Despite this new release, Mistral finds itself in catch up mode compared to bigger players. Its rivals OpenAI and Google have already released models with top scores in reasoning benchmarks, while Anthropic’s systems are recognized for advanced math and programming skills.
Current benchmarks show Magistral Medium doesn’t rise above Google’s Gemini 2.5 Pro or Anthropic’s Claude Opus 4 on tests that measure scientific and programming capabilities. On LiveCodeBench, a popular programming exam, Magistral Medium also comes up short behind those competitors.
Where Mistral believes Magistral stands out is its remarkable speed and the breadth of languages supported. According to the company, its chatbot delivers responses up to ten times faster than alternatives, and the models can converse smoothly in Italian, Arabic, Russian, and Simplified Chinese, among others.
Rather than pursuing outright dominance in technical performance, Mistral promotes Magistral as a tool for research, operations, and strategy. The model is designed to handle risk calculations, supply chain scenarios, and other decision tasks that often require juggling multiple factors at once, all while offering a step by step rationale that businesses can follow.
Magistral’s debut comes soon after several other launches from Mistral, including a coding assistant called Mistral Code and enterprise features for Le Chat. These tools aim to streamline corporate workflows, connecting the company’s AI models to widely used services like Gmail and SharePoint, and even letting users build their own smart agents.
