Mistral says "Le Chonk" can challenge the best AI models

Mistral has introduced Mistral Large 4, popularly referred to as Le Chonk, a 1 trillion-parameter model the company says will compete with leading closed-weight systems from U.S. and Chinese labs. The model is being distributed as an open-weight release, meaning anyone can use and customize it; a preview is available now and Mistral plans to ship a final version by the end of the month. Unlike general-purpose launches, Mistral positioned Le Chonk with explicit optimizations for coding and cyberdefense, and for domain-specific tasks in manufacturing, finance and electrical engineering. Guillaume Lample, Mistral’s cofounder and chief scientist, framed the model as targeting areas that other labs may underweight, suggesting targeted domain improvements as a route to competitive performance. Mistral presents Le Chonk as the most capable open-weight model produced outside China and says its performance is “very, very close” to some proprietary counterparts. The company additionally claims the model was trained from scratch rather than derived via distillation—a technique in which a smaller model learns from a larger model’s outputs—that U.S. officials have cited in criticisms of some Chinese labs. Mistral argues that open-weight models already cost less to run because users only pay for the compute they consume, and that narrowing performance differences removes remaining incentives for businesses to choose closed models. The Paris-based firm monetizes its work by charging pay-as-you-go fees for running models on its cloud and by supplying engineering services to tune models for customer needs. The release comes as Mistral has scaled quickly: the company led a $3.3 billion funding round in September at a $24 billion valuation, its largest-ever European tech raise, and has seen reported earnings grow roughly 20-fold in the past year. Despite having less capital and compute than large U.S. labs, Mistral says it remains in the race to produce top-tier models.