Ex-OpenAI researcher: training a last-gen frontier model isn't hard

Former OpenAI researcher Will Depue argued on X that training a frontier model one generation behind is not difficult, since knowledge diffuses quickly and large GPU clusters are increasingly available in neutral countries. He said the real challenge lies only in training the model at the very frontier.

2026-09-07 ~ 2026-09-07 · 2 related posts

• OpenAI researcher: training an N-1 frontier model is easy, only the marginal frontier is hard — willdepue · 2026-09-07
• X debate: is training an N-1 frontier model really easy? Ex-OpenAI researchers clash — anpaure · 2026-09-07