Mistral ships 1T-parameter open-weight ML4, beating GLM 5.3 on human eval, hiring worldwide
GuillaumeLample · x · 2026-10-06
Mistral co-founder and chief scientist Guillaume Lample announced the public preview of Mistral Large 4 (aka le Chonk): a natively multimodal model with 1T total parameters and 49B active, claimed to be the strongest open-weight model from the US or Europe. It hits open-model SOTA on cybersecurity, finance and law workloads and surpasses closed frontier models on visual grounding; weights drop end of the month. In human evaluation it beats GLM 5.3 on STEM, CAD and finance, and ties on agentic coding. Mistral is also scaling the team with hiring across Paris, London, Zurich, Warsaw, the US and Montreal for frontier open-weight training and large-scale RL.
Related event: Mistral Launches Mistral Large 4, a 1T-Parameter Open-Weight Flagship(86 posts)→
More from Companies & People
- Sarvam resells DeepSeek in India, drawing backlash over AI sovereignty stance — teortaxesTex · 2026-11-21
- arXiv thanks independent analyst storm for internal publishing pipeline reports — tdietterich · 2026-11-21
- If enough B2B AI companies assemble, an F1 race spontaneously appears — DenehyXXL · 2026-11-21
- Max Planck School of Biomedical AI opens first PhD cohort: 25 fully funded positions — kmborgwardt · 2026-11-21
- OpenAI researcher hints at mystery model "Inkling" in babysitting tweet — yiwenyuan98 · 2026-11-21
- OpenClaw hosts Agent Templates Focus Week with community co-building — heyneighbor · 2026-11-21