DeepSeek V4.1

Size

552B + 196B Engram

Published

DeepSeek · Sep 2026

Line

New parts in the block · after DeepSeek-V3.2-Exp

View

Source

DeepSeek‑V4.1 technical report @ 2cba9e4 · no training code is publishedHow to read

Effort-conditioned reward

z_{b,j}\sim\pi_\theta(\cdot\mid x,b)

{\htmlClass{vl-stream-k}{k}}(b)={\htmlClass{vl-stream-k}{k_0}}\,e^{-(b-b_{\min})/\tau}

{\htmlClass{vl-stream-context}{R}}=r-\min\{C_{\max},{\htmlClass{vl-stream-k}{k}}(b)\,\ell/L_{\mathrm{norm}}\}

R[B, |𝔅|, M]Centre in the subgroup A[B, |𝔅|, M]solid band: what this stage trainsthe update that uses A is outside this plate

Group-relative advantage

tensor

How much better than its subgroup each response was. The update that uses it is not written in the report.

Sources1

Contextual evidence

§5.2.1: prompts are dispatched by “the GRPO group size”; the report writes no objective, clipping or KL term.

group · deepseek-ai/DeepSeek-V4.1-Flash / DeepSeek_V41_Tech_Report.pdf
Pinned revision 2cba9e42aa02 ↗ · p. 30