WTF Is Going On(Weights, Tools & Frameworks)

한국어
deepseek-ai/DeepSeek-V4-Flash
text-generation mit FP8

What this is

291B parameters. Writes text from a prompt. This is the kind of model people mean by "an LLM".

Packaged as FP8. FP8 needs a GPU that supports it — Hopper or newer.

mit — commercial use allowed.

Release chainonly the steps this archive can prove

  1. weights posted2026-04-22

Measured herefrom our own daily capture — upstream publishes today only · one day in this window was never captured

Worth noticingmeasured from what we have collected — every number is recalculated and checked before it appears

← the board