OpenAI
availableShows if the model has enough results for an index.GPT-OSS 20B
GPT-OSS 20B is a non-reasoning model from OpenAI in the GPT-OSS family. 22 benchmarks count toward its score, in 7 categories.
IndexOverall score. 50 is the middle.40.5 ±4.4
CoverageShare of the index weight with results.90%
SpeedOutput tokens per second.222/s
Input / 1MUS dollars per 1M input tokens.$0.03
Output / 1MUS dollars per 1M output tokens.$0.13
ContextMaximum tokens in one request.131K
EloLMArena rating and rank.1287 (#246)
50 is the middle of the board. The range shows the doubt in the index.
10,393 votes. Elo shows what people prefer. It does not change the score.
CapabilitiesScore per category. 50 is the middle.
50 is the middleResults
22 counted| BenchmarkThe test name. | CategoryThe capability that the test measures. | ResultThe score from the publisher. | IndexThis result on the index scale. | RunThe settings of the run. | DateDate of the result. | Published byThe source of the result. |
|---|---|---|---|---|---|---|
| MATH 500 | Math | 94.2% | 47.1 | — | 9 Jan 2026 | Vals AI |
| MGSM | Multilingual | 89.0% | — | — | 9 Jan 2026 | Vals AI |
| AIME | Math | 86.0% | 53.8 | — | 16 Apr 2026 | Vals AI |
| LiveCodeBench | Coding | 80.4% | 57.4 | — | 1 Sept 2026 | Vals AI |
| MMLU Pro | Knowledge | 71.6% | 33.3 | — | 1 Sept 2026 | Vals AI |
| React Native Evals | Coding | 71.0% | 47.1 | — | — | Callstack |
| GPQA Diamond | Knowledge | 68.9% | 41.9 | — | 1 Sept 2026 | Vals AI |
| Artificial Analysis GPQA Diamond | Knowledge | 68.8% | 39.1 | — | — | Artificial Analysis |
| Artificial Analysis IFBench | Instruction | 65.1% | 55.2 | — | — | Artificial Analysis |
| τ²-Bench Tool-Agent-User Evaluation | Agentic | 60.2% | 42.0 | — | — | Victor Barres et al. |
| EuroEval Swedish | Multilingual | 55.5% | 71.6 | medium effort | — | EuroEval |
| EuroEval French | Multilingual | 52.9% | 68.3 | medium effort | — | EuroEval |
| EuroEval Swedish | Multilingual | 52.0% | 67.3 | low effort | — | EuroEval |
| EuroEval French | Multilingual | 51.7% | 66.8 | low effort | — | EuroEval |
| EuroEval Italian | Multilingual | 51.0% | 66.0 | medium effort | — | EuroEval |
| OTIS Mock AIME 2024-2025 | Math | 50.8% | 40.0 | high effort | — | Epoch AI |
| EuroEval Portuguese | Multilingual | 50.6% | 65.5 | medium effort | — | EuroEval |
| EuroEval Polish | Multilingual | 50.2% | 64.9 | medium effort | — | EuroEval |
| EuroEval Dutch | Multilingual | 49.9% | 64.6 | low effort | — | EuroEval |
| EuroEval Portuguese | Multilingual | 49.7% | 64.3 | low effort | — | EuroEval |
| EuroEval Polish | Multilingual | 48.6% | 62.9 | low effort | — | EuroEval |
| EuroEval Swedish | Multilingual | 48.4% | 62.7 | high effort | — | EuroEval |
| EuroEval Italian | Multilingual | 47.8% | 61.9 | low effort | — | EuroEval |
| GPQA diamond | Knowledge | 46.0% | 20.7 | high effort | — | Epoch AI |
| EuroEval Spanish | Multilingual | 45.8% | 59.5 | medium effort | — | EuroEval |
| EuroEval German | Multilingual | 45.8% | 59.4 | medium effort | — | EuroEval |
| EuroEval French | Multilingual | 44.8% | 58.2 | high effort | — | EuroEval |
| EuroEval Spanish | Multilingual | 44.2% | 57.5 | low effort | — | EuroEval |
| EuroEval Polish | Multilingual | 43.5% | 56.6 | high effort | — | EuroEval |
| EuroEval Italian | Multilingual | 43.3% | 56.4 | high effort | — | EuroEval |
| EuroEval German | Multilingual | 42.5% | 55.3 | low effort | — | EuroEval |
| EuroEval Portuguese | Multilingual | 40.7% | 53.1 | high effort | — | EuroEval |
| EuroEval Dutch | Multilingual | 40.5% | 52.9 | high effort | — | EuroEval |
| Artificial Analysis SciCode | Coding | 38.9% | 46.9 | — | — | Artificial Analysis |
| EuroEval German | Multilingual | 38.3% | 50.1 | high effort | — | EuroEval |
| Artificial Analysis Long Context Reasoning | Reasoning | 34.7% | 32.3 | — | — | Artificial Analysis |
| EuroEval Spanish | Multilingual | 32.9% | 43.5 | high effort | — | EuroEval |
| Artificial Analysis Coding Index | Coding | 20.7% | 33.6 | — | — | Artificial Analysis |
| Artificial Analysis Omniscience Accuracy | Knowledge | 16.0% | 33.7 | — | — | Artificial Analysis |
| Artificial Analysis Humanity's Last Exam | Knowledge | 11.0% | 36.5 | — | — | Artificial Analysis |
| Artificial Analysis Intelligence Index | Knowledge | 9.0% | 33.6 | — | — | Artificial Analysis |
| Artificial Analysis Agentic Index | Agentic | 1.4% | 38.3 | — | — | Artificial Analysis |
| Critical Physics Tasks | Reasoning | 1.4% | 41.3 | — | — | Artificial Analysis |
| APEX-Agents-AA | Agentic | 0.7% | 41.6 | — | — | Artificial Analysis / Mercor |
| GDPval-AA normalized | Agentic | 0.0% | 35.6 | — | — | Artificial Analysis |
| Chess Puzzles | Reasoning | 0.0% | 23.5 | high effort | — | Epoch AI |
22 benchmarks count, from 45 of 46 results. A grey row does not count. Too few models took that benchmark.