benchgap
OpenAI · model

GPT-OSS 20B benchmark scores

As of 2026-11-21, GPT-OSS 20B (OpenAI) has measured scores on 19 benchmarks and estimated scores on 80 more.

BenchmarkScoreSource
AA-GPQA Diamond68.8%measured
AA-HLE11.0%measured
AA-SciCode38.9%measured
AIME2690.9%estimated ± 4.3 pp, medium confidence
Artificial Analysis Intelligence Index9.8%estimated ± 3.2 pp, medium confidence
CritPt1.4%measured
GDPval-AA0.0%measured
GPQA Diamond71.7%estimated ± 1.2 pp, high confidence
HLE10.9%measured
IFBench63.7%estimated ± 6.6 pp, medium confidence
AA-LCR34.7%measured
LiveCodeBench v671.9%estimated ± 4.7 pp, high confidence
AA-Omniscience Accuracy16.5%estimated ± 2.4 pp, high confidence
SciCode39.1%estimated ± 3.1 pp, high confidence
BrowseComp8.5%estimated ± 4.2 pp, medium confidence
HLE w/ tools26.8%estimated ± 4.1 pp, low confidence
IFEval92.5%estimated ± 2.8 pp, high confidence
LongBench v251.3%estimated ± 5.5 pp, medium confidence
τ²-bench results60.2%measured
AA Coding Index20.7%measured
APEX-Agents0.0%estimated ± 9.4 pp, low confidence
APEX-Agents-AA0.7%measured
BioMysteryBench (human-difficult)20.9%estimated ± 6.8 pp, low confidence
DeepSearchQA26.7%estimated ± 4.5 pp, medium confidence
SWE-bench Verified60.7%measured
BFCL v453.9%estimated ± 2.7 pp, high confidence
CyberGym46.6%estimated ± 5.3 pp, low confidence
JobBench0.0%estimated ± 8.2 pp, low confidence
SWE-bench Pro37.1%estimated ± 4.5 pp, high confidence
WideResearch0.0%estimated ± 7.8 pp, low confidence
AA Agentic Index1.4%measured
MMLU-Pro72.3%estimated ± 4.0 pp, high confidence
AA-IFBench65.1%measured
AIME 202492.1%measured
BBH90.2%estimated ± 2.4 pp, low confidence
FailureSensorIQ43.5%measured
GPQA71.5%measured
MMLU85.3%measured
Vals GPQA Diamond67.9%estimated ± 3.2 pp, high confidence
Vals LiveCodeBench75.8%estimated ± 5.7 pp, low confidence
Vals MMLU-Pro79.6%estimated ± 2.3 pp, high confidence
MATH-50097.9%estimated ± 1.2 pp, medium confidence
MMMLU86.8%estimated ± 0.7 pp, medium confidence
AIME 202591.7%measured
AA MMLU-Pro85.3%estimated ± 1.2 pp, low confidence
Gert Labs33.6%estimated ± 8.3 pp, medium confidence
AA Harvey LAB v1.079.4%estimated ± 1.2 pp, low confidence
AA Tau3 Banking16.8%estimated ± 3.7 pp, medium confidence
ARC-AGI-124.7%estimated ± 6.7 pp, low confidence
ARC-AGI-20.0%estimated ± 10.9 pp, low confidence
CursorBench 3.236.0%estimated ± 5.3 pp, low confidence
FrontierCode 1.1 Main0.0%estimated ± 6.6 pp, low confidence
FrontierSWE v24.6%estimated ± 14.2 pp, low confidence
MazeBench0.4%estimated ± 5.5 pp, low confidence
OSWorld-Verified55.4%estimated ± 8.4 pp, low confidence
Vals SWE-bench44.6%estimated ± 5.5 pp, medium confidence
CWE-Bench-AA0.0%estimated ± 10.1 pp, low confidence
AA Cyber Index1.0%estimated ± 9.2 pp, low confidence
DeepsecBench-AA0.0%estimated ± 8.9 pp, low confidence
AA ITBench33.3%estimated ± 4.0 pp, medium confidence
Bug Hunt Bench13.9%estimated ± 10.5 pp, low confidence
CursorBench 4.025.6%estimated ± 4.9 pp, medium confidence
CWE-bench v10.0%estimated ± 6.4 pp, low confidence
DeepSWE28.7%estimated ± 7.5 pp, low confidence
DRACO0.0%estimated ± 2.0 pp, low confidence
FrontierCode 2.026.1%estimated ± 9.3 pp, low confidence
HLE w/o tools11.5%estimated ± 5.3 pp, medium confidence
PostTrainBench v1.125.3%estimated ± 7.4 pp, low confidence
Toolathlon-Verified5.5%estimated ± 2.3 pp, low confidence
VulcanBench Frontier v468.5%estimated ± 6.8 pp, low confidence
GDP.pdf17.0%estimated ± 4.6 pp, medium confidence
ARC-AGI-31.3%estimated ± 9.8 pp, low confidence
ARC-AGI-3 Provider Adapter2.7%estimated ± 7.3 pp, low confidence
Vibe Code Bench27.7%estimated ± 6.5 pp, low confidence
Hermes Bench0.0%estimated ± 2.9 pp, medium confidence
Hermes Index0.0%estimated ± 8.7 pp, low confidence
C-Eval92.2%estimated ± 1.1 pp, medium confidence
HMMT Feb 202594.9%estimated ± 2.9 pp, high confidence
HMMT Feb 202676.2%estimated ± 11.0 pp, low confidence
MCP Atlas64.6%estimated ± 7.8 pp, low confidence
MMLU-Redux90.9%estimated ± 1.7 pp, high confidence
NL2Repo12.1%estimated ± 4.7 pp, medium confidence
QwenClawBench55.9%estimated ± 6.0 pp, low confidence
SuperGPQA52.5%estimated ± 7.2 pp, medium confidence
Toolathlon27.0%estimated ± 6.3 pp, low confidence
AA LiveCodeBench84.0%estimated ± 4.7 pp, low confidence
HealthBench Hard12.3%estimated ± 8.6 pp, low confidence
MMLU-Pro (Arcee)82.4%estimated ± 4.0 pp, medium confidence
React Native Evals71.0%measured
ResearchClawBench2.8%estimated ± 1.9 pp, medium confidence
SWE-Rebench57.0%estimated ± 3.6 pp, low confidence
BioMysteryBench (human-solvable)81.7%estimated ± 0.6 pp, low confidence
ScarfBench24.4%estimated ± 9.1 pp, low confidence
SimpleQA22.9%estimated ± 7.8 pp, medium confidence
LiveCodeBench66.5%estimated ± 14.6 pp, low confidence
OpenHarmony Bench v1.051.0%estimated ± 2.4 pp, medium confidence
MRCRv222.8%estimated ± 11.3 pp, low confidence
MMLongBench-Doc57.3%estimated ± 2.6 pp, low confidence
BioBasicsBench33.9%estimated ± 10.7 pp, low confidence