Mistral announces Large 4 as a planned trillion-parameter open-weight flagship

Mistral's Large 4, also called Le Chonk or ML4, is described as a sparse multimodal model with roughly 1T total parameters, 49B active parameters, 160+ languages, and a 4,000-GPU training footprint. Limited preview access is available, with weights planned for October 27; preliminary coding signals are promising, but final artifacts, licensing, safety documentation, detailed benchmarks, and independent validation remain pending.