Mistral AI’s October 6, 2026 launch of Mistral Large 4 (“Le Chonk”), a 1.05-trillion-parameter multimodal MoE model with 49–52 billion active parameters, sets the immediate context for the next release (4.1+). The model entered public preview via Mistral Studio with a 1M-token context window and strong reported results on cybersecurity, finance, manufacturing, coding, and visual-grounding benchmarks, outperforming other US and European open-weight systems while closing gaps with leading Chinese open models. Trained from scratch on roughly 3,800–4,000 NVIDIA Grace Blackwell GPUs in Mistral’s European data centers, it emphasizes data sovereignty and efficient inference. Open weights are scheduled for late October (around Oct 27–31), after safety testing with partners and governments, followed by specialized derivatives. This rapid iteration cadence, backed by a €3 billion Series D and partnerships such as ASML and Samsung, reinforces trader expectations that Mistral remains the primary driver for its own subsequent flagship releases amid competition from DeepSeek, Qwen, and closed US labs.