Llama 3.1 405B Instruct — Benchmarks

Benchmark scores for Llama 3.1 405B Instruct aggregated from public leaderboards, with how it ranks among open models. See hardware requirements for what you need to run it.

Overall rank: #52 of 78 open modelscomposite 38.1/100 across 7 benchmarks in 4 categories · methodology

Coding

BenchmarkScoreOpen rankAll models
WeirdML21.4%#40 / 46#146 / 161

Instruction Following

BenchmarkScoreOpen rankAll models
LMArena Text1335.2#81 / 182#220 / 391

Knowledge

BenchmarkScoreOpen rankAll models
MMLU-Pro73.3%#34 / 144#90 / 257
MMLU84.5%#6 / 85#13 / 136

Math

BenchmarkScoreOpen rankAll models
MATH Level 549.8%#15 / 39#64 / 107
AIME 2024/20259.7%#56 / 82#220 / 270

Reasoning

BenchmarkScoreOpen rankAll models
DTBench61.4%#41 / 67#161 / 210
GPQA Diamond50.9%#51 / 94#212 / 289
SimpleBench23.0%#20 / 25#90 / 101
WinoGrande82.2%#7 / 51#13 / 79

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