Mistral Medium 3.5 Model Card

This document describes Mistral AI's Mistral Medium 3.5 model: what it can do, how to access it, and what it costs. The model weights are publicly available under a Modified MIT license, meaning you can use, modify, and redistribute them within the terms of that license. If you use the hosted API, Mistral AI charges $1.50 per million input tokens and $7.50 per million output tokens.

This model card establishes the technical specifications, API surface, pricing structure, and licensing terms for Mistral AI's Mistral Medium 3.5 model. Mistral AI makes the model weights publicly available as open weights under a Modified MIT license, governing permitted use, modification, and redistribution. The model operates with a 256,000-token context window and is accessible via a set of defined API endpoints covering Chat Completions, Chat Moderations, Moderations, Function Calling, Agents, Conversations, and Built-In Tools. Mistral AI prices input tokens at $1.50 per one million tokens and output tokens at $7.50 per one million tokens, with the governing legal terms taking effect April 28, 2026.

For an individual user, this document establishes two things that directly affect usage. First, the model can process up to 256,000 tokens in a single request, setting a firm ceiling on how much text or conversation history can be submitted at once. Second, costs accrue at $1.50 per million input tokens and $7.50 per million output tokens, meaning output generation is five times more expensive than input on a per-token basis. Users who wish to work with the model weights directly may do so under the Modified MIT license, which defines what is legally permitted in terms of use, modification, and redistribution.