Mistral Large 4

Not yet independently verified. Only Mistral’s own announcement of 6 October 2026 has been read; no independent dated report is cited yet. Context window and licence are not stated, and the weights are promised for the end of October, so this listing covers the preview API. We will update this when it can be confirmed, and remove this note.

Mistral Large 4 is a mixture-of-experts model from Mistral AI with about one trillion parameters in total and 49 billion active per token, released as a public preview API on Mistral Studio on 6 October 2026. Mistral describes it as a hybrid instruct-and-reasoning model with multimodal input that handles instruction following, reasoning and agentic work in one model, and says it was trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in its own European data centres, where the preview is also served. The announcement page lists the preview at $1.36 per million input tokens and $4.18 per million output tokens. Mistral says it will release the weights by the end of October and is red-teaming the model with cybersecurity firms, vetted partners and state authorities until then; the context window and the licence for the weights are not stated yet. The capability claims in the post, including cybersecurity, finance and manufacturing results, are Mistral’s own evaluations, so no score is stored here. For buyers who need models hosted in Europe or want the option of running weights themselves, it is the first frontier-scale model from Mistral since its earlier Large releases, and the weights date is the one to watch.