Mistral AI’s October 6, 2026 preview launch of Mistral Large 4 (“Le Chonk”), a 1T-parameter natively multimodal MoE model with ~49-52B active parameters, drives current trader sentiment on the next 4.1+ release. The model, trained from scratch on ~3,800 NVIDIA Grace Blackwell GPUs in European data centers, claims state-of-the-art open-weight results in cybersecurity, finance, coding, and visual grounding while supporting 160+ languages. Weights are scheduled for late October (targets cited around October 27-31), following safety red-teaming with partners and governments. This rapid follow-up to prior releases, backed by a €3B Series D and plans for specialized derivatives, signals continued fast iteration and competitive pressure against Chinese open models and U.S. closed frontier systems. Key near-term catalysts include the weights drop and any post-release benchmark updates or roadmap details that could accelerate a 4.1 milestone.