FP4

Quartet: Native FP4 Training Can Be Optimal for Large Language Models
We introduce a new approach and scaling law for accurate, end-to-end FP4 training.
Andrei Panferov
Feb 25, 2026 1 min read