Nathan Lambert: Cohere Model Trained via Narrow RL on Internal Infrastructure, Bad for Open Weights

AI researcher Nathan Lambert explained that Cohere's open model was likely the product of narrow RL training reliant on Cohere's internal infrastructure, a design he argues is unfavorable for open-weight models.

2026-10-27 ~ 2026-10-27 · 2 related posts

• Nat Lambert: Cohere's model is narrow RL training on internal infra ill-suited for open weights — natolambert · 2026-10-27
• Nat Lambert: narrow RL training on internal infra is a poor fit for open-weight models — mervenoyann · 2026-10-27