Mistral Large 4 'Le Chonk' preview priced at $1.36/M input; open weights due October 31

Large 4 adds a 1.6B vision encoder, a 1M-token context window and support for 160+ languages. Mistral trained it on its own cluster of 3,800 Grace Blackwell GPUs.

Simon Willison flagged a limitation: the API exposes only two reasoning levels, "none" and "high." The model also trails Claude Opus 5 and GPT-5.5 on Terminal-Bench and Humanity's Last Exam.

Mistral is promoting STEMBench wins in engineering and astronomy. Its co-founder is also urging Europe to invest in its own AI companies.