The company launched a preview of Large 4 on October 6, ahead of a planned open-weight release.

On October 6, 2026, Mistral AI opened a public API preview of Mistral Large 4, nicknamed “Le Chonk,” and said its weights would arrive by month’s end. The multimodal mixture-of-experts model has 1.05 trillion total parameters, about 49 billion active per token, and a 1.6-billion-parameter vision encoder; Mistral reports a one-million-token context window. It scored 61.7% on DeepSWE v1.1, 59.4% on SWE-Atlas-QnA and 28.3% on Terminal-Bench 4, targeting developers building coding agents and enterprises seeking models they can eventually run on their own infrastructure. That prospect matters in a market where Chinese open-weight systems have set fierce competition: Mistral presents Large 4 as a bid to narrow the gap, though preview access is not the same as downloadable weights. The company says it trained the model from scratch on 3,800 Nvidia Grace Blackwell GPUs in European data centres. Large 4 follows a roughly five-month gap between major releases; Mistral’s prior flagship launched in December 2025, with an update in April 2026.