MistralLaunchAI ModelsOctober 6, 2026

Mistral launches Mistral Large 4 "Le Chonk" in public preview

Read original source →mistral.ai

The 1.05T-parameter MoE (49B active per token) is natively multimodal with a 1M-token context, trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral's European datacenters. API is live at $1.36/M input and $4.18/M output; weights drop end of October.

How this story unfolded

2 weeks · 7 reports · 5 community posts · 12 of 17 shown

  1. Oct 6
  2. Oct 7
  3. Oct 21

More stories today

Open the live feed