Mistral Large 4 Preview

Mistral Large 4 (Public Preview; aka ML4 / le Chonk) is Mistral AI's flagship general-purpose multimodal Mixture-of-Experts model. Mistral's docs describe it as 1.05T total parameters with 52B active parameters and a 1.6B vision encoder, while Mistral's Hugging Face upcoming-release page describes it as a 1T model with 49B active parameters per token (52B including embeddings and output layers). It supports a 1M-token context window with text+image input and text output, plus structured outputs, function calling, document QnA, batching, agents/conversations, and built-in tools. Launch API pricing is 50% off for two weeks versus list ($0.68/$0.07/$2.09 sale vs $1.36/$0.14/$4.18 list per million input/cached/output tokens). Weights are planned for the end of October 2026 and are not yet released, so the catalog keeps the model proprietary until a public weights license is published.