Mistral releases Le Chonk, a 1T-parameter open model targeting coding and cyberdefense

Guillaume Lample, Mistral’s cofounder and chief scientist, told WIRED that Le Chonk fills gaps left by other labs, which often prioritize general-purpose models. Mistral emphasizes it was trained from scratch—unlike Chinese models accused of using distillation to mimic proprietary AI. Open-weight models like Le Chonk cut costs for businesses, as they only require compute resources, not licensing fees. Mistral argues this reduces hesitation around open-source adoption.

Despite its progress, Mistral still trails OpenAI and Anthropic in performance, revenue, and release frequency. The French lab funds operations through cloud-based pay-as-you-go fees and engineering services. Its recent $3.3 billion funding round at a $24 billion valuation reflects its growth, with earnings reportedly rising 20-fold over the past year.