Mistral Large 4
Mistral Large 4 is Mistral AI's new flagship, a roughly 1-trillion-parameter mixture-of-experts model that handles text and images and offers both quick answers and step-by-step reasoning. It launched on October 6, 2026 as a public preview API, with open weights…
Best for
Teams that want a frontier-class multimodal model from a European vendor with self-hosting and sovereign deployment options.
Overview
Mistral Large 4 is Mistral AI's new flagship, a roughly 1-trillion-parameter mixture-of-experts model that handles text and images and offers both quick answers and step-by-step reasoning. It launched on October 6, 2026 as a public preview API, with open weights promised by the end of October.
Strengths and limits
- About 1T total / 52B active parameter MoE with hybrid instruct and reasoning
- Strong reported cybersecurity and agentic results (Cybench 93%)
- 1M-token context and promised open weights for self-hosting
- Still a public preview; behavior may change and weights were not yet published at launch
- Ranked behind Claude Opus 5 in Mistral's own cited human coding eval
Published benchmark scores
Scores are as published by the source; settings (reasoning effort, tools, attempts) differ between labs, so compare with care. How to read benchmarks →