Mistral previews Large 4, a 1-trillion-parameter model with weights due by end of October

Mistral AI opened a public preview of Mistral Large 4 on its Mistral Studio API, a natively multimodal model with 1 trillion parameters and 52 billion active, trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral’s own European data centres on data spanning more than 160 languages. Mistral says the weights will be released by the end of the month and that until then it is red-teaming the model with cybersecurity leaders, vetted partners and state authorities, who get a version with reduced moderation and expanded cyber capabilities. The company says the model ranks in the top five on the Artificial Analysis Cyber Index, scores 82% on a test that asks a model to reproduce and patch a real open-source vulnerability, solves 93% of Cybench, and scores 61.7% on DeepSWE v1.1 and 28.3% on Terminal-Bench 4. It argues that closed models such as Claude Opus 5.5 and GPT-6 Astra score near zero on the vulnerability test because they refuse it, and that defenders need open weights they can run under their own policies. A European deployment is operated by Mistral end to end under European law.

When

6 Oct 2026

ID

WTA-N-2026-0328

Where

France, European Union

Who

Mistral AI, NVIDIA

Topics

Model releases, Open-weight models, Cybersecurity, Sovereign AI, Frontier models

Next stop on this Storyline: LinkedMicrosoft says enterprise AI agents need Guardian Agents to watch every action19 Oct 2026

Spotted an error? Email [email protected].