Mistral Large 4, a 1.05-trillion-parameter multimodal MoE model with 49-52 billion active parameters and a 1M-token context window, entered public API preview on October 6, 2026, marking the lab’s strongest open-weight release to date and the first major milestone from its €3 billion Series D round. The preview emphasizes enterprise strengths in cybersecurity, coding, finance, and agentic tasks, where it outperforms many Chinese open-weight rivals and approaches some closed frontier models in targeted benchmarks. Official statements indicate open weights, architecture details, and post-training methodology will drop by the end of October (target dates cited as October 27–31), after safety reviews with partners and authorities. Mistral has signaled rapid iteration ahead, positioning Large 4 as the foundation for specialized follow-on models and noting ongoing training runs that could deliver sizable capability gains within weeks. This compressed timeline, combined with Europe-focused sovereignty messaging and competitive pressure from DeepSeek, Qwen, and U.S. labs, shapes trader views on when a 4.1+ update is likely to surface.