DeepSeek v3 0324 — Benchmarks

Benchmark scores for DeepSeek v3 0324 aggregated from public leaderboards, with how it ranks among open models. See hardware requirements for what you need to run it.

Overall rank: #20 of 73 open modelscomposite 56.6/100 across 7 benchmarks in 4 categories · methodology

Coding

BenchmarkScoreOpen rankAll models
SWE-bench Verified42.0%#12 / 13#117 / 163
Aider Polyglot55.1%#8 / 18#33 / 69

Knowledge

BenchmarkScoreOpen rankAll models
MMLU-Pro81.3%#18 / 119#57 / 259

Math

BenchmarkScoreOpen rankAll models
AIME 2024/202537.8%#14 / 34#96 / 155
MATH Level 575.5%#5 / 32#40 / 108

Reasoning

BenchmarkScoreOpen rankAll models
SimpleBench27.2%#14 / 19#73 / 90
GPQA Diamond67.6%#14 / 46#91 / 182

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