V4.1 Is the First Model With a Numerical Reasoning Effort Parameter Tied to Length Penalty

nrehiew_ · x · 2026-09-11

nrehiew observes that V4.1 is the first model he has seen with a numerical reasoning effort parameter that directly influences the length penalty — letting users numerically trade off reasoning depth against output verbosity.

Related event: DeepSeek V4.1 Tech Report Deep Dive: RL Infrastructure, Sandbox Design and Inference Stack(8 posts)→

Original post →

More from Models

Models channel →