Mistral Large 4 open weights ship with 1M context

Mistral’s decision to release Mistral Large 4 as open weights with a 1M-token context window reflects a broader push toward models that can handle long documents and agentic workflows. The model is specifically aimed at agentic coding and cybersecurity, and Mistral is offering customization through Mistral Forge, which lets teams adapt the model to their own data and tasks. The pricing — $0.70 per 1M input tokens and $2.10 per 1M output tokens, about 50% below the Preview — suggests Mistral is trying to make a large-context model more accessible for production use. However, the MEOW license includes a clause requiring a separate commercial license for model-as-a-service businesses, which could affect cloud providers that want to offer the model as a hosted service. The release also includes a 256K maximum output length, which may be useful for generating long code or reports in a single pass.