| Benchmark | Claude Opus 4.8 | GPT-4o |
|---|---|---|
| AA Agentic Index | 49.4 | 9.7 |
| AA Intelligence | 57.3 | 11.2 |
| AA-LCR | 73 | 55 |
| AA-Omniscience | 28.8 | -10.5 |
| arc_agi_1 | 92.5 | 4.5 |
| ARC-AGI-2 | 72.1 | 0 |
| arena_vision | 1280 | 1162 |
| Artificial Analysis Coding Index | 74.3 | 24.2 |
| critpt | 20.9 | 0 |
| Fortress | 18.2 | 47.2 |
| gdpval | 54.2 | 0 |
| GPQA Diamond | 93.6 | 70.1 |
| HLE | 57.9 | 5.3 |
| IFBench | 62.2 | 36 |
| longbench_v2 | 69.1 | 48.1 |
| mathvision | 86.7 | 30.4 |
| MMMU-Pro | 78.9 | 59.9 |
| OmniScience Accuracy | 48.8 | 23.7 |
| OmniScience Non-Hallucination | 60.7 | 62.1 |
| OSWorld-Verified | 83.4 | 5 |
| scicode | 53.5 | 33.4 |
| simplebench | 64.8 | 17.8 |
| SWE-bench Verified | 88.6 | 38.8 |
| Terminal-Bench Hard | 58.3 | 8.3 |
| Video-MME | 86 | 77.2 |
| τ²-Bench Telecom (AA run) | 94.4 | 28.9 |
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.