DeepSeek V4 Pro 0813 — Benchmarks

Benchmark scores for DeepSeek V4 Pro 0813 aggregated from public leaderboards, with how it ranks among open models. See hardware requirements for what you need to run it.

Overall rank: #15 of 123 open modelscomposite 67/100 across 10 benchmarks in 4 categories · methodology

Coding

BenchmarkScoreOpen rankAll models
SciCode49.2%#8 / 50#63 / 155
LiveBench Coding77.2#6 / 16#29 / 54

Knowledge

BenchmarkScoreOpen rankAll models
SimpleQA52.9%#1 / 15#16 / 78

Math

BenchmarkScoreOpen rankAll models
AIME 2024/202598.6%#1 / 74#12 / 270
LiveBench Math95.1#1 / 16#8 / 54
ProofBench50.0%#3 / 21#18 / 64

Reasoning

BenchmarkScoreOpen rankAll models
GPQA Diamond91.7%#3 / 83#22 / 291
ARC-AGI90.5%#2 / 16#47 / 200
LiveBench Reasoning85.8#4 / 16#25 / 54
CritPt18.0%#4 / 50#41 / 167

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