DeepSeek V4 Flash — Benchmarks

Benchmark scores for DeepSeek V4 Flash aggregated from public leaderboards, with how it ranks among open models. See hardware requirements for what you need to run it.

Overall rank: #29 of 123 open modelscomposite 59.4/100 across 6 benchmarks in 3 categories · methodology

Coding

BenchmarkScoreOpen rankAll models
SciCode44.9%#17 / 50#86 / 155
LiveBench Coding69.2#15 / 16#50 / 54

Math

BenchmarkScoreOpen rankAll models
LiveBench Math79.7#14 / 16#48 / 54

Reasoning

BenchmarkScoreOpen rankAll models
LiveBench Reasoning70.6#15 / 16#52 / 54
SimpleBench46.3%#11 / 23#60 / 101
CritPt7.1%#11 / 50#74 / 167

Scores aggregated from public benchmark sources (each linked from the benchmark pages). llmrun does not run these benchmarks.