OpenAI / MODEL EVIDENCE
22 benchmarks · 31 observations · 24 sources · released 2025-08-05
22 benchmarks · large dot: gpt-oss-20b; small dots: closest models where reported
GPT OSS 20B · GPQA Diamond · Accuracy ↗
60.8% · gpt-oss-20b · GPQA diamond · Best score (across scorers) · % · Epoch AI · GPQA diamond ↗
53.16% · gpt-oss-20b · GPQA diamond · Best score (across scorers) · % · Epoch AI · GPQA diamond ↗
45.96% · gpt-oss-20b · GPQA diamond · Best score (across scorers) · % · Epoch AI · GPQA diamond ↗
gpt-oss-20b · OTIS Mock AIME 2024-2025 · Best score (across scorers) · % ↗
50.83% · gpt-oss-20b · OTIS Mock AIME 2024-2025 · Best score (across scorers) · % · Epoch AI · OTIS Mock AIME 2024-2025 ↗
40.28% · gpt-oss-20b · OTIS Mock AIME 2024-2025 · Best score (across scorers) · % · Epoch AI · OTIS Mock AIME 2024-2025 ↗
gpt-oss-20b · Terminal Bench · Accuracy mean · % ↗
3.07% · gpt-oss-20b · Terminal Bench · Accuracy mean · % · Epoch AI · Terminal Bench ↗
3.4% · gpt-oss-20b · Terminal Bench · Accuracy mean · % · Epoch AI · Terminal Bench ↗
3.1% · gpt-oss-20b · Terminal Bench · Accuracy mean · % · Epoch AI · Terminal Bench ↗
gpt-oss-20b · WeirdML · Accuracy · % ↗
36.83% · gpt-oss-20b · WeirdML · Accuracy · % · Epoch AI · WeirdML ↗
gpt-oss-20b · CritPt · Accuracy · % ↗
gpt-oss-20b · LMCA · Score · % ↗
gpt-oss-20b · DTBench · Accuracy · % ↗
gpt-oss-20b · SciCode · Score · % ↗
gpt-oss-20b · General ECI · 26 October 2026 fit ↗
GPT OSS 20B · LegalBench · Accuracy ↗
GPT OSS 20B · CaseLaw · Accuracy ↗
GPT OSS 20B · CorpFin · Accuracy ↗
GPT OSS 20B · TaxEval · Accuracy ↗
GPT OSS 20B · MedQA · Accuracy ↗
GPT OSS 20B · AIME · Accuracy ↗
GPT OSS 20B · MATH 500 · Accuracy ↗
GPT OSS 20B · MGSM · Accuracy ↗
GPT OSS 20B · MMLU Pro · Accuracy ↗
GPT OSS 20B · LiveCodeBench · Accuracy ↗
gpt-oss-20b · Scale Labs · Risk elicitation score ↗
gpt-oss-20b · Scale Labs · Score ↗
gpt-oss-20b · Average contest performance ↗
Similar positions on shared benchmark leaderboards. Smaller gaps are closer.
9 of 13 shared benchmark comparisons
9 of 13 shared benchmark comparisons
9 of 13 shared benchmark comparisons
Mean absolute gap in percentile positions on shared benchmark settings; lower is closer. At least 60% overlap is required, with a minimum of three when available. Independent evaluations are used when there are at least three; otherwise published report comparisons are used. Repeated model observations within one setting are averaged before ranking. Writing-style traits and composite indices are excluded. This describes similarity, not an overall quality ranking.
Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.
GPT OSS 20B · gpt-oss-20b