gpt-oss-120b (high): benchmark scores, price and ranking
gpt-oss-120b (high) from OpenAI ranks #447 of 743 LLMs on the LLMs Tiger aggregate benchmark score (47.3), based on 10 benchmark results captured up to 2026-10-13.
- Organisation
- OpenAI
- Type
- Open-weight / local model
- Open weights
- Yes
- Released
- 2025-08-05 (source: curated)
- Cheapest API price
- $0.030 in / $0.170 out per million tokens (via wandb, observed 2026-10-13T19:47Z)
- Aggregate rank
- #447 of 743 (score 47.3)
Relative to other models, gpt-oss-120b (high) is strongest on OTIS Mock AIME (#46 of 189) and weakest on Mystery Games (#59 of 59).
Across the 10 benchmarks where at least ten models are scored, gpt-oss-120b (high) is in the top 10% on 0 and in the top half on 5.
Its cheapest listed input price, $0.030 per million tokens, is lower than that of 96% of the 427 models with a listed price.
gpt-oss-120b (high) benchmark results
| Benchmark | Score | Rank | Source | Captured |
|---|---|---|---|---|
| OTIS Mock AIME | 88.9% | #46 of 189 | Epoch AI Benchmarking Hub | 2026-10-13 |
| WeirdML | 48.2% | #32 of 109 | Epoch AI Benchmarking Hub | 2026-10-13 |
| Chess Puzzles | 20.0% | #45 of 122 | Epoch AI Benchmarking Hub | 2026-10-13 |
| DTBench | 76.3% | #64 of 146 | Epoch AI Benchmarking Hub | 2026-10-13 |
| GPQA Diamond | 75.8% | #125 of 275 | Epoch AI Benchmarking Hub | 2026-10-13 |
| CritPt | 1.1% | #67 of 115 | Epoch AI Benchmarking Hub | 2026-10-13 |
| LMCA | 22.1% | #70 of 108 | Epoch AI Benchmarking Hub | 2026-10-13 |
| SciCode | 34.0% | #85 of 106 | Epoch AI Benchmarking Hub | 2026-10-13 |
| Fiction.liveBench | 44.4% | #40 of 47 | Epoch AI Benchmarking Hub | 2026-10-13 |
| Mystery Games | 0.0% | #59 of 59 | Epoch AI Benchmarking Hub | 2026-10-13 |
About gpt-oss-120b (high)
gpt-oss-120b (high) is an open-weight model that can run on your own hardware from OpenAI. It was released on 2025-08-05, according to the developer's announcement. The cheapest API listing we track charges $0.030 per million input tokens and $0.170 per million output tokens, through wandb. We hold 10 benchmark results for it across 4 areas, and none is in the top tenth of its board.
Where gpt-oss-120b (high) stands
On our aggregate score, gpt-oss-120b (high) is #447 of 743 ranked models, in the lower half. Its score is 47.3; the scale is centred on 50, so it sits 2.7 points below the typical model. That beats 40% of the ranked field. Among the 88 ranked OpenAI models it is #56. The next one up is GPT-5.1 (Thinking, None) at 47.5. The next one down is gpt-5.6-terra-xhigh (codex-harness) at 47.3.
gpt-oss-120b (high), benchmark by benchmark
OTIS Mock AIME puts gpt-oss-120b (high) at 88.9%, #46 of 189. That beats 76% of the models tested, but 11.1 points short of Claude Fable 5.1 (max) (100.0%). The results either side are GPT-5.4 mini (xhigh) (88.9%) and Inkling (xhigh) (88.9%). It covers competition maths (AIME 2024-25 style). The board median is 57.8%, so gpt-oss-120b (high) is 31.1 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
On WeirdML, gpt-oss-120b (high) reaches 48.2%. It ranks #32 of 109, ahead of 71% of the field, and 45.1 points behind GPT-6 Astra (max) at 93.3%. The results either side are GLM-5 (48.2%) and o1-preview (47.6%). It covers unusual machine-learning tasks. The board median is 41.2%, so gpt-oss-120b (high) is 7.0 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
gpt-oss-120b (high) manages 20.0% on Chess Puzzles, good for #45 of 122. GPT-6 Astra (max) leads at 72.0%, 52.0 points further up, and 63% of models sit below. The results either side are DeepSeek v4 Pro (max) (20.0%) and Kimi K2 Thinking Turbo (20.0%). It covers chess tactics puzzles. The board median is 12.5%, so gpt-oss-120b (high) is 7.5 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
DTBench puts gpt-oss-120b (high) at 76.3%, #64 of 146. That beats 56% of the models tested, but 22.7 points short of Claude Opus 5.5 (max) (98.9%). The results either side are gemini-2.5-flash (76.5%) and qwen3-235b-a22b (75.7%). It covers dynamic tool-use benchmark. The board median is 68.8%, so gpt-oss-120b (high) is 7.5 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
On GPQA Diamond, gpt-oss-120b (high) reaches 75.8%. It ranks #125 of 275, ahead of 55% of the field, and 20.0 points behind GPT-6 Astra (max) at 95.8%. The results either side are Gemma 4 31B IT (75.8%) and grok-3-beta (75.8%). It covers expert-level science questions. The board median is 71.7%, so gpt-oss-120b (high) is 4.0 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
gpt-oss-120b (high) lands below the middle on CritPt: 1.1%, #67 of 115, with 57% of the field scoring higher. GPT-5.6 Sol (max) leads at 32.3%; the closest result above is GLM-4.6 at 1.1%. It covers unpublished research-level physics problems. The board median is 2.9%, so gpt-oss-120b (high) is 1.7 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
gpt-oss-120b (high) lands below the middle on LMCA: 22.1%, #70 of 108, with 64% of the field scoring higher. Claude Opus 5.5 (max) leads at 68.2%; the closest result above is o1 (high) at 22.3%. It covers language-model capability assessment. The board median is 29.2%, so gpt-oss-120b (high) is 7.1 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
SciCode is a weak spot. gpt-oss-120b (high) scores 34.0%, #85 of 106, behind 79% of the models measured. The leader, Claude Opus 5.5 (max), reaches 66.9%. It covers scientific research code. The board median is 44.3%, so gpt-oss-120b (high) is 10.3 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
Fiction.liveBench is a weak spot. gpt-oss-120b (high) scores 44.4%, #40 of 47, behind 83% of the models measured. The leader, grok-4-0709, reaches 94.4%. It covers long-context story comprehension (16k). The board median is 61.1%, so gpt-oss-120b (high) is 16.7 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
Mystery Games is a weak spot. gpt-oss-120b (high) scores 0.0%, #59 of 59, behind 98% of the models measured. The leader, GPT-6 Astra (max), reaches 84.0%. It covers deductive reasoning puzzles. The board median is 23.0%, so gpt-oss-120b (high) is 23.0 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
Strengths and weaknesses
Reasoning, maths and science: solid, ahead of 57% of the field on average across 3 benchmarks. Its best board here is OTIS Mock AIME (#46 of 189) and its weakest is CritPt (#67 of 115). General knowledge and chat: uneven, ahead of 37% of the field on average across 5 benchmarks. Its best board here is WeirdML (#32 of 109) and its weakest is Mystery Games (#59 of 59).
Price and value
The median input price across the 427 priced models is $0.500 per million tokens, so gpt-oss-120b (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 Qwen3.8-27B (XHigh) ($0.040) and Mercury 2.5 (high) ($0.040).
Release history
The OpenAI model released before it was o4-mini (high) on 2025-04-16, which scores 46.8. The next one was gpt-oss-20b (high) on 2025-08-05, at 44.3.
Reading these numbers
The 10 scores on this page come from 1 source: Epoch AI Benchmarking Hub (10). No official leaderboard supplied any of them, so treat them as third-party measurements. The leaderboard also lists 4 other reasoning-effort settings of this model, which redirect here. This page shows the highest.
gpt-oss-120b (high) is also listed at 4 other reasoning-effort settings; this page shows the highest-effort row, which is the one the leaderboard keeps by default.
Models ranked nearby
- #444 Kimi K2.7 Code (47.3)
- #445 command-a-03-2025-quality (47.3)
- #446 Cogito v2.1 (671B) (47.3)
- #448 Kimi K2 0905 (Novita) (47.3)
- #449 gpt-5.6-terra-xhigh (codex-harness) (47.3)
- #450 Llama-3.2-3B-Instruct (47.2)
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