Mistral’s first reasoning model, Magistral, launches with large and small Apache 2.0 version
1 min read
Summary
European AI company, Mistral, has launched Magistral, a large language model (LLM) family, which marks its entrance into the AI “reasoning” space.
The family consists of two releases: a proprietary version for enterprise customers, and an open-source version (Magistral Small) to reinforce the company’s commitment to an open-source ecosystem.
The open-source version will allow developers to build and deploy apps using Mistral’s latest reasoning architecture for free, without licensing fees or vendor lock-in.
Benchmarking tests show that it outperforms competitors such as Deepseek across a range of tests, including mathematical accuracy, question-answering, and coding challenges.
Key differentiators for Magistral are said to be its chain-of-thought logic, multilingual capabilities, and speed, making it ideal for use across a range of use cases in law, finance, healthcare, software architecture, and data engineering, as well as creative writing and storytelling.