TL;DR
On October 6, 2026, French AI company Mistral AI launched a public preview of its flagship model Mistral Large 4, nicknamed 'Le Chonk': a MoE architecture with 1.05 trillion total and 49 billion active parameters, natively multimodal, with open weights expected by the end of October (per Mistral's official release).
Event overview
On October 6, 2026, French AI company Mistral AI launched a public preview of its flagship model Mistral Large 4, nicknamed 'Le Chonk', with the preview API available through Mistral Studio. Per MarkTechPost, Unite.AI and AI News citing Mistral's official information, the model uses a MoE architecture with 1.05 trillion total and 49 billion active parameters, is natively multimodal, and its open weights are expected by the end of October.
Key facts
- Event: Mistral Large 4 public preview (October 6, 2026, per Mistral's official release).
- Specs: 1.05T total and 49B active parameters; granular MoE; natively multimodal (with a 1.6B vision encoder); 1M-token context window (per AI/TLDR citing Mistral's official docs).
- Training: trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral's own EU datacenters (per MarkTechPost).
- Open plan: open weights expected by the end of October; currently in public preview (per AI News and Unite.AI).
- Performance claim: Mistral says Large 4 is among the strongest open-weight models from Europe and the US and rivals the strongest Chinese open systems (per Anadolu Agency and The Decoder citing Mistral, a company claim).
- Benchmark: scores 61.7% on DeepSWE v1.1 (per AI/TLDR citing Mistral's official docs).
Official sources
- MarkTechPost (October 6, 2026): Mistral AI Releases Mistral Large 4 (Le Chonk): A 1.05T Parameter Multimodal MoE Model
- Unite.AI (October 6, 2026): Mistral Unveils 1.05T-Parameter Le Chonk MoE Model in Public Preview
- AI News (October 6, 2026): Mistral AI launches Large 4 preview ahead of open-weight release