Mistral Medium is a proprietary large language model developed by Mistral AI. It was first introduced in December 2023 as part of the company's tiered offering of performance-oriented models. Positioned between the open-weight Mixtral 8x7B and the flagship Mistral Large, it is designed to provide a balance of high-level reasoning capabilities and low operational latency.
The model is optimized for multilingual tasks and professional use cases, supporting languages such as English, French, German, Spanish, and Italian. It is particularly effective for intermediate reasoning tasks, including data extraction, document summarization, and code generation. Unlike Mistral's open-weight models, Mistral Medium is a closed-source model accessible via API services.
Technical Details
Mistral Medium initially launched with a 32,768 token context window. In later iterations, such as the 2025 update to Mistral Medium 3.1, this window was expanded to 128,000 tokens. While Mistral AI has not officially disclosed the exact parameter count or architecture of the model, it is built on a transformer-based framework and is evaluated as a high-performance alternative to other leading proprietary and open-source systems.