PyyanThe AI Index

Pyyan / Open-Weight Models / DeepSeek V4.1 Flash

DeepSeek V4.1 Flash

Twice the parameters of the Flash it replaces, and cheaper to run.

currentDeepSeek ↗Verified 2026-09-13

90.6%

Terminal-Bench 2.1

Specifications

Context1Msource · as of 2026-09-13
Input ($/Mtok)$0.3source · as of 2026-09-13 · medium confidence
Output ($/Mtok)$1.2source · as of 2026-09-13 · medium confidence
LicenceMITsource · as of 2026-09-13
Max output256Ksource · as of 2026-09-13
Parameters552B total, 8B / 16B activesource · as of 2026-09-13
Released10 Sep 2026source · as of 2026-09-13
Model IDdeepseek-flashsource · as of 2026-09-13

Official links

DeepSeek V4.1 Flash model pageDeepSeek official site

What matters

  • 552B total, 8B active while reading the prompt and 16B while generating, with a 1M token context
  • DeepSeek says it beats V4-Pro; V4-Pro requests route to it from 14 September 2026
  • MIT licence, native image input, and a KV cache about a quarter the size of V4-Flash's

Best for cheap agentic coding with images.

Compare with

All 5 side by side →

DeepSeek V4.1 Flash vs Qwen3.8-Flash-Next →DeepSeek V4.1 Flash vs DeepSeek V4-Pro →DeepSeek V4.1 Flash vs Atria Dawn Preview →DeepSeek V4.1 Flash vs DeepSeek V4 Flash Vision Exp →DeepSeek V4.1 Flash vs Inkling →

Also in Open-Weight Models

Qwen3.8-Flash-Next · AlibabaDeepSeek V4-Pro · DeepSeekAtria Dawn Preview · Shanghai AI LaboratoryDeepSeek V4 Flash Vision Exp · DeepSeekInkling · Thinking Machines Lab