DeepSeek v2 — Benchmarks

Benchmark scores for DeepSeek v2 aggregated from public leaderboards, with how it ranks among open models. See hardware requirements for what you need to run it.

Knowledge

Benchmark	Score	Open rank	All models
HellaSwag	87.1%	#4 / 42	#6 / 76
MMLU	78.4%	#14 / 76	#33 / 136

Reasoning

Benchmark	Score	Open rank	All models
BIG-Bench Hard	78.8%	#5 / 37	#8 / 50

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