Model

gpt-oss-20b

OpenAI · open weights

openai.com

Knowledge & Reasoning

Knowledge & Reasoning
BenchmarkScoreMetricHarnessSourceConfidenceLink
CritPt1.4accuracyN/AIndependentlowepoch.ai
GeoFitBench-LLM72.7success_rateN/AIndependentmediumarxiv.org
GPQA Diamond60.8accuracyN/AIndependenthighepoch.ai
OTel61.4scoreN/AIndependentmediumarxiv.org

Math

Math
BenchmarkScoreMetricHarnessSourceConfidenceLink
Prim43.4generationN/AIndependentmediumarxiv.org

Coding

Coding
BenchmarkScoreMetricHarnessSourceConfidenceLink
AlgoREval88.0accuracyN/AIndependentmediumarxiv.org
AptMQL-Bench51.0soft-exN/AIndependentmediumarxiv.org
SciCode34.4scoreN/AIndependentmediumartificialanalysis.ai

Agentic / Tool Use

Agentic / Tool Use
BenchmarkScoreMetricHarnessSourceConfidenceLink
DelegationBench50.6accuracyN/AIndependentmediumarxiv.org
DrugTargetWorld7.4scoreN/AIndependentmediumarxiv.org
MAGMA-Bench40.4success_rateN/AIndependentmediumarxiv.org
ReFract0.3success_ratefunction callingIndependentmediumarxiv.org
Terminal-Bench3.4accuracyMini-SWE-AgentIndependentmediumtbench.ai
Terminal-Bench3.1accuracyTerminus 2Independentmediumtbench.ai

Multimodal

Multimodal
BenchmarkScoreMetricHarnessSourceConfidenceLink
VisualNoiseQA42.2accuracybpIndependentmediumarxiv.org

Safety & Alignment

Safety & Alignment
BenchmarkScoreMetricHarnessSourceConfidenceLink
ObligationBench9.7recallN/AIndependentmediumarxiv.org

Abstract Reasoning

Abstract Reasoning
BenchmarkScoreMetricHarnessSourceConfidenceLink
ArgGYM65.7accuracyN/AIndependentmediumarxiv.org
auto-nesy-bench70.3f1zsIndependentmediumarxiv.org

Reproduced by the community

Loading community runs…