June 11 News.FranceArtificial Intelligence Laboratory Mistral announced its entry into the field of inferential AI models on Tuesday.On June 10, Mistral officially launched its firstinference modelseries - Magistral. This series of models aims to improve consistency and reliability in subject areas such as math and physics by solving problems in steps, similar to other inference models such as OpenAI's o3 model and Google's Gemini 2.5 Pro.
The Magistral family consists of two versions: Magistral Small and Magistral Medium. Magistral Small, with 24 billion parameters, is available for download under the Apache 2.0 license on the AI development platform Hugging Face. Magistral Medium, a more capable model, is currently in preview and available only on Mistral's Le Chat chat platform, company APIs, and third-party partner cloud platforms.
In its blog post, Mistral says, "Magistral is suitable for a wide range of enterprise-class application scenarios, including structured computation, procedural logic, decision trees, and rule-based systems. These models are fine-tuned with multi-step logic that enhances interpretability and provides a traceable thought process in the language of the user."
Founded in 2023, Mistral is a cutting-edge modeling lab working on a range of AI-powered services, including Le Chat and mobile apps. The company is backed by venture capital firms such as General Catalyst and has raised more than €1.1 billion (note: about Rs. 9,022 million at current exchange rates) in funding.
Despite its strong resources, Mistral lags behind other leading AI labs in some areas, such as developing inference models. Looking at Mistral's own benchmarks, Magistral's competitiveness also appears to fall short. In the GPQA Diamond and AIME tests, which assess the physics, math, and science skills of models, Magistral Medium did not perform as well as Gemini 2.5 Pro and Anthropic's Claude Opus 4. In addition, in the popular programming benchmark LiveCodeBench, Magistral Medium failed to outperform Gemini 2.5 Pro. Medium also fails to outperform the Gemini 2.5 Pro in the popular programming benchmark LiveCodeBench.
Perhaps that's why Mistral highlighted other advantages of Magistral in its blog post, claiming that Magistral answers questions on Le Chat 10 times faster than competitors and supports multiple languages, including Italian, Arabic, Russian and Simplified Chinese.
In the article, the company writes, "Based on our flagship model, Magistral is designed for research, strategic planning, operational optimization, and data-driven decision making, whether it's performing multi-factor risk assessment and modeling or calculating the optimal delivery window under constraints."
Magistral's release comes shortly after Mistral launched Mistral Code, its "ambient programming" client. And a few weeks ago, Mistral also released several programming-focused models and launched Le Chat Enterprise, an enterprise-focused chat service that offers tools such as an AI agent builder and integrates Mistral's models with third-party services such as Gmail and SharePoint.