Inference costs dominate training costs when serving models to billions of users or trillions of tokens, and within training itself, reinforcement learning (which is a wrapper on inference) is starting to exceed the compute requirements of pre-training.
factualpending
Speaker
TanishkEvidence Quote
“not only are inference costs dominating training costs, but even within training, RL is starting to exceed the compute requirements of pre-training”
Created: 8/12/2026, 6:25:34 PM
Your notes
Loading notes...