Report
Mistral’s Le Chonk model and the Western open-weight AI race
The interviewer says his talk with Mistral research lead Pierre Stock covers training Le Chonk on 3,800 GPUs.
TLDR
The video description identifies Le Chonk as Mistral Large 4, a one-trillion-parameter open-weight model. The interviewer’s chapter list covers mixture-of-experts, reinforcement learning, open weights versus cyber refusals, and a preview of Mistral Large 5.
Combined views
3.1K
4 Sources, first seen ago
