Overview

• Magistral Small, a 24-billion parameter model under the Apache 2.0 license, is available for free download on Hugging Face
• Magistral Medium is offered at $2 per million input tokens and $5 per million output tokens through Mistral’s API, Le Chat platform and partner clouds including Amazon SageMaker
• Both models generate traceable chain-of-thought outputs and handle complex tasks such as mathematics and coding across multiple languages including French, Spanish, Arabic and Simplified Chinese
• Benchmark tests show Magistral Medium matching or exceeding predecessor scores on AIME-24 and GPQA Diamond while delivering up to ten times faster token throughput in Mistral’s Le Chat environment
• Backed by Microsoft and valued at $6.2 billion, Mistral’s dual-release approach bolsters its position against U.S. and Chinese AI rivals by combining open-source community engagement with enterprise revenue opportunities