Mistral AI introduced Mistral Large 4 on October 6 as its new flagship model for reasoning, coding, document understanding and agentic work. The company says the sparse mixture-of-experts architecture contains 1 trillion total parameters while activating 49 billion for each token.
Public preview now, open weights later
The model is available in public preview through Mistral's API and Le Chat. Mistral says downloadable weights are planned for release by the end of October, so this announcement is an API and hosted-product launch rather than a completed open-weight release.
Mistral reports native text-and-image input, a 1 million-token context window and improved performance across coding and multimodal evaluations. Those benchmark and efficiency claims are vendor-reported; no independent reproduction was located during this review.