OpenAI
availableShows if the model has enough results for an index.o3-mini
o3-mini is a reasoning model from OpenAI in the o3 family. 23 benchmarks count toward its score, in 6 categories.
IndexOverall score. 50 is the middle.40.3 ±4.9
CoverageShare of the index weight with results.85%
SpeedOutput tokens per second.160/s
Input / 1MUS dollars per 1M input tokens.$1.1
Output / 1MUS dollars per 1M output tokens.$4.4
ContextMaximum tokens in one request.200K
EloLMArena rating and rank.1319 (#221)
50 is the middle of the board. The range shows the doubt in the index.
56,655 votes. Elo shows what people prefer. It does not change the score.
CapabilitiesScore per category. 50 is the middle.
50 is the middleResults
23 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 level 5 | Math | 96.5% | 46.1 | high effort | — | Epoch AI |
| Instruction-Following Eval | Instruction | 93.9% | 53.6 | — | — | Jeffrey Zhou et al. |
| MATH 500 | Math | 91.8% | 44.4 | — | 9 Jan 2026 | Vals AI |
| MGSM | Multilingual | 91.3% | — | — | 9 Jan 2026 | Vals AI |
| American Invitational Mathematics Examination 2024 | Math | 87.3% | — | — | — | Mathematical Association of America |
| Massive Multitask Language Understanding | Knowledge | 86.9% | — | — | — | Dan Hendrycks et al. |
| AIME | Math | 86.5% | 54.0 | — | 16 Apr 2026 | Vals AI |
| MMLU Pro | Knowledge | 78.7% | 44.5 | — | 1 Sept 2026 | Vals AI |
| Graduate-Level Google-Proof Q&A | Knowledge | 77.2% | 49.5 | — | — | David Rein et al. |
| GPQA diamond | Knowledge | 77.0% | 49.3 | high effort | — | Epoch AI |
| OTIS Mock AIME 2024-2025 | Math | 76.9% | 54.6 | high effort | — | Epoch AI |
| GPQA Diamond | Knowledge | 75.5% | 48.0 | — | 1 Sept 2026 | Vals AI |
| Artificial Analysis GPQA Diamond | Knowledge | 74.8% | 45.2 | — | — | Artificial Analysis |
| LiveCodeBench | Coding | 71.5% | 49.3 | — | 1 Sept 2026 | Vals AI |
| Software Engineering Benchmark Verified | Coding | 49.3% | 37.1 | — | — | Carlos E. Jimenez et al. |
| SWE-bench Verified | Coding | 42.4% | 31.5 | Agentless Lite | 26 Feb 2026 | SWE-bench team |
| SWE-bench Verified | Coding | 42.4% | 31.5 | Agentless Lite | 1 Sept 2026 | SWE-bench team |
| ARC-AGI-1 (semi-private) | Reasoning | 34.5% | 43.5 | high effort | — | ARC Prize Foundation |
| SWE-bench Lite | Coding | 32.3% | 30.8 | Agentless Lite | 11 Sept 2025 | SWE-bench team |
| SWE-bench Lite | Coding | 31.3% | 30.0 | Aegis | 11 Sept 2025 | SWE-bench team |
| τ²-Bench Tool-Agent-User Evaluation | Agentic | 28.7% | 19.4 | — | — | Victor Barres et al. |
| FrontierMath-Tiers-1-3-v2-Private | Math | 18.6% | 41.9 | high effort | — | Epoch AI |
| Chess Puzzles | Reasoning | 17.0% | 45.5 | high effort | — | Epoch AI |
| SimpleQA Verified | Knowledge | 15.3% | 35.7 | high effort | — | Epoch AI |
| Artificial Analysis Intelligence Index | Knowledge | 12.5% | 38.0 | — | — | Artificial Analysis |
| FrontierMath-2025-02-28-Private | Math | 12.4% | 42.2 | high effort | — | Epoch AI |
| Artificial Analysis Humanity's Last Exam | Knowledge | 7.9% | 33.1 | — | — | Artificial Analysis |
| Mystery Game Puzzles | Reasoning | 7.0% | 41.4 | high effort | — | Epoch AI |
| FrontierMath-Tier-4-2025-07-01-Private | Math | 4.2% | 48.0 | high effort | — | Epoch AI |
| ARC-AGI-2 (semi-private) | Reasoning | 3.0% | 41.2 | high effort | — | ARC Prize Foundation |
| FrontierMath-Tier-4-v2-Private | Math | 0.0% | 47.9 | high effort | — | Epoch AI |
23 benchmarks count, from 28 of 31 results. A grey row does not count. Too few models took that benchmark.