Mistral Large 4, a 1.05-trillion-parameter multimodal MoE model with 49-52 billion active parameters, entered public preview via Mistral Studio on October 6, 2026, after training on 3,800 Nvidia Grace Blackwell GPUs in European data centers. The company positions it as state-of-the-art among open-weight models on cybersecurity, finance, manufacturing, and visual-grounding benchmarks while closing gaps with frontier closed systems in select areas. Open weights are scheduled for release around October 27 following safety testing with partners and governments. This rapid iteration—following Large 3 in late 2025—reflects Mistral’s €3 billion Series D funding, emphasis on European AI sovereignty, and competition from Chinese open-weight labs plus U.S. closed models. Traders are watching the weights rollout and any roadmap signals for clues on the cadence toward a 4.1+ successor.