gpt-oss-20b (high): benchmark scores, price and ranking
gpt-oss-20b (high) from OpenAI ranks #644 of 743 LLMs on the LLMs Tiger aggregate benchmark score (44.3), based on 8 benchmark results captured up to 2026-10-13.
- Organisation
- OpenAI
- Type
- Open-weight / local model
- Open weights
- Yes
- Released
- 2025-08-05 (source: epoch)
- Cheapest API price
- $0.015 in / $0.070 out per million tokens (via darkbloom, observed 2026-10-13T19:47Z)
- Aggregate rank
- #644 of 743 (score 44.3)
Relative to other models, gpt-oss-20b (high) is strongest on DTBench (#76 of 146) and weakest on Chess Puzzles (#108 of 122).
Across the 8 benchmarks where at least ten models are scored, gpt-oss-20b (high) is in the top 10% on 0 and in the top half on 0.
Its cheapest listed input price, $0.015 per million tokens, is lower than that of 99% of the 427 models with a listed price.
gpt-oss-20b (high) benchmark results
| Benchmark | Score | Rank | Source | Captured |
|---|---|---|---|---|
| DTBench | 68.0% | #76 of 146 | Epoch AI Benchmarking Hub | 2026-10-13 |
| WeirdML | 40.9% | #58 of 109 | Epoch AI Benchmarking Hub | 2026-10-13 |
| CritPt | 1.4% | #64 of 115 | Epoch AI Benchmarking Hub | 2026-10-13 |
| OTIS Mock AIME | 50.8% | #106 of 189 | Epoch AI Benchmarking Hub | 2026-10-13 |
| GPQA Diamond | 46.0% | #206 of 275 | Epoch AI Benchmarking Hub | 2026-10-13 |
| SciCode | 34.4% | #84 of 106 | Epoch AI Benchmarking Hub | 2026-10-13 |
| LMCA | 14.5% | #89 of 108 | Epoch AI Benchmarking Hub | 2026-10-13 |
| Chess Puzzles | 0.0% | #108 of 122 | Epoch AI Benchmarking Hub | 2026-10-13 |
About gpt-oss-20b (high)
gpt-oss-20b (high) is an open-weight model that can run on your own hardware from OpenAI. It was released on 2025-08-05, according to Epoch AI. The cheapest API listing we track charges $0.015 per million input tokens and $0.070 per million output tokens, through darkbloom. We hold 8 benchmark results for it across 4 areas, and none is in the top tenth of its board.
Where gpt-oss-20b (high) stands
On our aggregate score, gpt-oss-20b (high) is #644 of 743 ranked models, in the lower half. Its score is 44.3; the scale is centred on 50, so it sits 5.7 points below the typical model. That beats 13% of the ranked field. Among the 88 ranked OpenAI models it is #76. The next one up is GPT-4o (May 2024) at 44.3. The next one down is o3-mini (high) at 44.0.
gpt-oss-20b (high), benchmark by benchmark
gpt-oss-20b (high) lands below the middle on DTBench: 68.0%, #76 of 146, with 51% of the field scoring higher. Claude Opus 5.5 (max) leads at 98.9%; the closest result above is GPT-4.1 at 68.3%. It covers dynamic tool-use benchmark. The board median is 68.8%, so gpt-oss-20b (high) is 0.8 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
gpt-oss-20b (high) lands below the middle on WeirdML: 40.9%, #58 of 109, with 52% of the field scoring higher. GPT-6 Astra (max) leads at 93.3%; the closest result above is gemini-2.5-flash-preview-05-20_16K at 41.0%. It covers unusual machine-learning tasks. The board median is 41.2%, so gpt-oss-20b (high) is 0.2 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
gpt-oss-20b (high) lands below the middle on CritPt: 1.4%, #64 of 115, with 55% of the field scoring higher. GPT-5.6 Sol (max) leads at 32.3%; the closest result above is Gemma 4 31B IT at 1.4%. It covers unpublished research-level physics problems. The board median is 2.9%, so gpt-oss-20b (high) is 1.4 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
gpt-oss-20b (high) lands below the middle on OTIS Mock AIME: 50.8%, #106 of 189, with 56% of the field scoring higher. Claude Fable 5.1 (max) leads at 100.0%; the closest result above is DeepSeek-R1-Distill-Llama-70B at 51.4%. It covers competition maths (AIME 2024-25 style). The board median is 57.8%, so gpt-oss-20b (high) is 6.9 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
gpt-oss-20b (high) lands below the middle on GPQA Diamond: 46.0%, #206 of 275, with 75% of the field scoring higher. GPT-6 Astra (max) leads at 95.8%; the closest result above is Llama-3.2-90B-Vision-Instruct at 46.1%. It covers expert-level science questions. The board median is 71.7%, so gpt-oss-20b (high) is 25.8 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
gpt-oss-20b (high) lands below the middle on SciCode: 34.4%, #84 of 106, with 78% of the field scoring higher. Claude Opus 5.5 (max) leads at 66.9%; the closest result above is Magistral Small 1.2 at 35.2%. It covers scientific research code. The board median is 44.3%, so gpt-oss-20b (high) is 10.0 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
LMCA is a weak spot. gpt-oss-20b (high) scores 14.5%, #89 of 108, behind 81% of the models measured. The leader, Claude Opus 5.5 (max), reaches 68.2%. It covers language-model capability assessment. The board median is 29.2%, so gpt-oss-20b (high) is 14.7 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
Chess Puzzles is a weak spot. gpt-oss-20b (high) scores 0.0%, #108 of 122, behind 88% of the models measured. The leader, GPT-6 Astra (max), reaches 72.0%. It covers chess tactics puzzles. The board median is 12.5%, so gpt-oss-20b (high) is 12.5 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
Strengths and weaknesses
Reasoning, maths and science: uneven, ahead of 38% of the field on average across 3 benchmarks. Its best board here is CritPt (#64 of 115) and its weakest is GPQA Diamond (#206 of 275). General knowledge and chat: weak, ahead of 25% of the field on average across 3 benchmarks. Its best board here is WeirdML (#58 of 109) and its weakest is Chess Puzzles (#108 of 122).
Price and value
The median input price across the 427 priced models is $0.500 per million tokens, so gpt-oss-20b (high) costs less than the typical listing. Models that score within 1.5 points or better and cost less: Qwen2.5-Coder-7B-Instruct ($0.010, 49.5), DeepSeek V4.1 Flash (high) ($0.010, 48.9), Qwen2.5-Coder-3B-Instruct ($0.010, 46.5). Similar scores cost more on granite-4.0-micro ($0.017) and Mistral-Nemo-Instruct-2407 ($0.019).
Release history
The OpenAI model released before it was gpt-oss-120b (high) on 2025-08-05, which scores 47.3. The next one was gpt-5-chat on 2025-08-07, at 49.2.
Reading these numbers
The 8 scores on this page come from 1 source: Epoch AI Benchmarking Hub (8). No official leaderboard supplied any of them, so treat them as third-party measurements. The leaderboard also lists 3 other reasoning-effort settings of this model, which redirect here. This page shows the highest.
gpt-oss-20b (high) is also listed at 3 other reasoning-effort settings; this page shows the highest-effort row, which is the one the leaderboard keeps by default.
Models ranked nearby
- #641 grok-code-fast-1 (44.3)
- #642 Qwen1.5-7B-Chat (44.3)
- #643 GPT-4o (May 2024) (44.3)
- #645 Yi-6B-Chat (44.3)
- #646 Claude 3.5 Sonnet (Jun 2024) (44.2)
- #647 DeepSeek-V3 (Mar 2025) (44.1)
More from OpenAI
- GPT-6 Astra (max) (#2, 64.0)
- GPT-6.1 Sol (max) (#5, 60.6)
- GPT-5.6 Sol (max) (#6, 60.1)
- GPT-6 Sol (max) (#7, 60.0)
- GPT-5.5 (xhigh) (#11, 58.0)
- GPT-5.6 Terra (max) (#12, 57.8)
Compare it on the live leaderboard → · How the aggregate score works