Ripple-65B Final Run: 12T Tokens, Roughly 9x Overtraining
Ripple-65B's final run used 12T tokens for a 65B-parameter model, about 185 tokens per parameter or roughly 9x overtraining, while still following the power law.
2026-11-06 ~ 2026-11-06 · 2 related posts
- Ripple-65B final run: 12T tokens for 65B params, roughly 9x overtraining — _katieeverett · 2026-11-06
- Ricursive 65B Model Trained on 12T Tokens With ~9x Overtraining, Power Law Holds — _katieeverett · 2026-11-06