DeepSeek V4-Flash

DeepSeek llm

Smaller, faster sibling of V4-Pro sharing the same 1M-token architecture. Scores are from its Think Max reasoning mode.

overall ⌄

57.9

39 of 125 ranked

price · $/M tokens

0.14 / 0.28

in / out, as of 2026-07-25

to run it yourself ⌄

170 GB

minimum · 682 GB recommended

Benchmarks

What its maker published, and what anyone else measured. Every figure links the document it came from.

benchmarkpublishedmeasured
MMLU-Pro86.2 —
GPQA88.1 —
SWE-bench79 —
SWE-bench Pro52.6 —
HLE34.8 —
BrowseComp73.2 —

Cheaper, and at least as good

Nothing is both cheaper and ranked as high. This model is on the value frontier — whatever you pay for it, you are paying for something.

Lineage

view in full graph →

Metadata

orgDeepSeek
released2026-04-22
params284B-A13B
licensemit
hf iddeepseek-ai/DeepSeek-V4-Flash