| Benchmark | GPT-4o | o1 |
|---|---|---|
| AA Agentic Index | 9.7 | 31.1 |
| AA Intelligence | 11.2 | 24 |
| AA-LCR | 55 | 63.3 |
| AA-Omniscience | -10.5 | -10.6 |
| aider_polyglot | 30.7 | 61.7 |
| AIME 2024 | 9.3 | 79.2 |
| AIME 2025 | 11.7 | 81.7 |
| AIR-Bench 2024 | 62.4 | 80 |
| anthropic_red_team | 99.1 | 98.3 |
| arena_vision | 1162 | 1193 |
| Artificial Analysis Coding Index | 24.2 | 39.7 |
| bbq | 95.1 | 97.3 |
| Codeforces | 759 | 2061 |
| Codeforces (Percentile) | 23.6 | 96.6 |
| Codeforces (Rating) | 759 | 2061 |
| critpt | 0 | 0.3 |
| DROP (3-shot F1) | 83.7 | 90.2 |
| Fortress | 47.2 | 19.4 |
| gdpval | 0 | 11.5 |
| GPQA Diamond | 70.1 | 75.7 |
| GSM8K | 95.6 | 97.1 |
| harmbench | 82.9 | 96.3 |
| HLE | 5.3 | 7.7 |
| humaneval | 90.6 | 88.1 |
| IFBench | 36 | 70.3 |
| livecodebench_pass1cot | 34.2 | 63.4 |
| MATH | 85.3 | 96.4 |
| math_500_em | 74.6 | 96.4 |
| MathVista | 63.8 | 71.8 |
| metr_hcast | 25 | 42.6 |
| metr_re_bench | 0 | 0 |
| metr_swaa | 99.2 | 100 |
| mgsm | 90.5 | 90.8 |
| mmlu | 88.1 | 91.8 |
| mmmlu | 81.4 | 87.7 |
| MMMU | 72.2 | 77.6 |
| OmniScience Accuracy | 23.7 | 34.7 |
| OmniScience Non-Hallucination | 62.1 | 30.7 |
| scicode | 33.4 | 35.8 |
| SEAL VISTA | 34.9 | 45.3 |
| simple_safety_tests | 98.5 | 99 |
| simplebench | 17.8 | 41.7 |
| simpleqa | 39.4 | 47 |
| SWE-bench Verified | 38.8 | 48.9 |
| TAU-bench (airline) | 42.8 | 50 |
| TAU-bench (retail) | 60.3 | 70.8 |
| Terminal-Bench Hard | 8.3 | 12.9 |
| xstest | 97.3 | 97 |
| τ²-Bench Telecom (AA run) | 28.9 | 62.6 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.