OpenAI, released Aug 5, 2025
GPT OSS 20Bprice, context, benchmarks and release details
- Input, per 1M tokens
- not yet reported
- Output, per 1M tokens
- not yet reported
- Context window
- 131Kmodels.devPublished source fact
Retrieved Oct 9, 2026 · MIT
Open source ↗ - Max output
- 32.8Kmodels.devPublished source fact
Retrieved Oct 9, 2026 · MIT
Open source ↗ - Released
- Aug 5, 2025models.devPublished source fact
Retrieved Oct 9, 2026 · MIT
Open source ↗
How the score breaks down
Weights: reasoning 30%, math 15%, coding 40%, preference 15%. Results use fixed 0–100 scales before averaging, and thin evidence is pulled toward 50. Method si-v3-retained-evidence-2, computed Oct 9, 2026, 05:13 UTC.
Around it on the leaderboard
- 98 Step 3.5 Flash 53.0
- 99 Qwen3 30B A3B 52.6
- 100 GPT OSS 20B 52.5
- 101 Claude Sonnet 3.5 v2 51.9
- 102 Claude Opus 4.1 51.5
Benchmark results
3 benchmarks, 7 resultsEach row shows the best published result. Where a model was tested at several settings, such as reasoning effort, open the row to see each one. Hover or tap a value for its source.
GPQA Diamondreasoning 60.8%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] mediumPublished Aug 27, 2026
Retrieved Oct 9, 2026 · CC-BY
Open source ↗
60.8 3 settings
- low effort 53.2%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] lowPublished Aug 27, 2026
Retrieved Oct 9, 2026 · CC-BY
Open source ↗ - medium effort 60.8%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] mediumPublished Aug 27, 2026
Retrieved Oct 9, 2026 · CC-BY
Open source ↗ - high effort 46.0%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] highPublished Aug 6, 2026
Retrieved Oct 9, 2026 · CC-BY
Open source ↗
OTIS Mock AIME 2024–2025math 65.3%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] mediumPublished Aug 27, 2026
Retrieved Oct 9, 2026 · CC-BY
Open source ↗
65.3 3 settings
- low effort 40.3%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] lowPublished Aug 27, 2026
Retrieved Oct 9, 2026 · CC-BY
Open source ↗ - medium effort 65.3%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] mediumPublished Aug 27, 2026
Retrieved Oct 9, 2026 · CC-BY
Open source ↗ - high effort 50.8%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] highPublished Aug 27, 2026
Retrieved Oct 9, 2026 · CC-BY
Open source ↗
Open source ↗ 60.7
Normalization uses a fixed 0–100 scale for each unit, independent of other models. Compare evaluation conditions before reading a small gap as decisive. “Lab-reported” marks the provider's own published figure.
Details and sources
- Open weights
- Yesmodels.devPublished source fact
Retrieved Oct 9, 2026 · MIT
Open source ↗ - License
- not yet reported
- Input modalities
- textmodels.devPublished source fact
Retrieved Oct 9, 2026 · MIT
Open source ↗ - First seen by SuperIndex
- Oct 8, 2026
- Coverage
- 51% of expected source weight
Reported (2)
- Epoch AI BenchmarkingOct 8, 2026
- LMArena / ArenaOct 8, 2026
Awaiting (5)
- ARC Prize13% of weight
- Humanity’s Last Exam13% of weight
- Official model cards via models.dev4% of weight
- LiveBench13% of weight
- Terminal-Bench6% of weight
Confidence rises as pending sources publish. Some sources never cover some models, so confidence reaches 100% at 80% of expected weight.