Mistral's ML4 was trained on 3,800 Grace Blackwell GPUs in its European datacenter
BLUECOW009 · x · 2026-10-07
Mistral chief scientist Guillaume Lample revealed ML4 was trained on 3,800 NVIDIA Grace Blackwell GPUs at the company's Bruyères-le-Châtel datacenter in Europe, funded by its Series B. Series C and D clusters are coming online soon, enabling longer training, more ambitious post-training and faster iteration, with 'large and rapid improvements' expected in weeks and months. Observers joked they had to triple-check it wasn't a shitpost.
Related event: Mistral Trained ML4 on 3,800 Grace Blackwell GPUs in Europe(3 posts)→
More from Companies & People
- Blogger defends Jensen Huang against negative coverage, citing firsthand meetings — kimmonismus · 2026-11-22
- OpenAI alumni's Applied Compute acquired by Palo Alto Networks for $4.5B — martinvars · 2026-11-22
- Nadella wants AI tokens cheap and boring — betting against premium token pricing — VraserX · 2026-11-22
- Granola cofounder on why AI drowns knowledge workers and how to design a calm presence — soleio · 2026-11-22
- Anthropic locked down B2B while conceding B2C — irony in YC's tarpit advice — minhnguyen · 2026-11-22
- Nous Research, Pangram host sold-out NYC Non-Slop Hackathon with cash prizes — Teknium · 2026-11-22