Mistral Large 4

Mistral's open-weight 1T multimodal model for coding, agents, and defensive security.

Also known as Le Chonk, mistral-large-4

What it does

Mistral Large 4, also called Le Chonk, is Mistral's open-weight multimodal model. The October 29, 2026 announcement releases the weights after a public preview earlier that month. The Hugging Face model card describes a 1 trillion-parameter mixture-of-experts model with 49 billion active parameters per token, or 52 billion including embeddings and output layers, a 1 million token context window, and text and image input with text output. It lists 128 routed experts, 4 active per token, and 1 shared expert, and reasoning efforts of low, medium, and high, with high as the default. Mistral says it was trained from scratch in Europe. The API name is mistral-large-4. The announcement prices the API at $0.70 per million input tokens, $0.07 per million cached input tokens, and $2.10 per million output tokens. Weights use the Mistral Ecosystem Open-Weight License, which Mistral calls a modified MIT license. Mistral says you can use, modify, fine-tune, and commercialize the model, and that a Model-as-a-Service business needs a separate commercial license. On Mistral's numbers, the Artificial Analysis Coding Agent Index score is 48%, tied with GLM-5.3 while using 47% fewer tokens per task, Terminal-Bench 4.0 is 41%, and the Artificial Analysis Cyber Index is 51%. The model card lists an Artificial Analysis Intelligence Index v5 score of 26, the highest among western open-weight models, and says leading open-weight models from China score higher on that aggregate. The Hugging Face repository was created on September 30, 2026 and last modified on October 29, 2026.

Who is Mistral Large 4 for?

Developers and security teams who want an open-weight model they can self-host for coding and defensive work.

Tags

• open-weight
• coding
• multimodal