GPT OSS 120B — Benchmarks

Benchmark scores for GPT OSS 120B aggregated from public leaderboards, with how it ranks among open models. See hardware requirements for what you need to run it.

Overall rank: #36 of 123 open modelscomposite 57/100 across 10 benchmarks in 4 categories · methodology

Coding

BenchmarkScoreOpen rankAll models
Terminal-Bench18.7%#14 / 16#52 / 57
Aider Polyglot41.8%#10 / 18#45 / 69
SciCode38.9%#26 / 50#116 / 155
SWE-bench Verified26.0%#16 / 16#142 / 162
SWE-bench Bash Only26.0%#8 / 9#42 / 48

Knowledge

BenchmarkScoreOpen rankAll models
MMLU-Pro80.8%#21 / 119#61 / 259

Math

BenchmarkScoreOpen rankAll models
AIME 2024/202588.9%#13 / 74#63 / 270

Reasoning

BenchmarkScoreOpen rankAll models
GPQA Diamond75.8%#25 / 83#139 / 291
SimpleBench22.1%#21 / 23#94 / 101
CritPt1.1%#25 / 50#108 / 167

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