| Benchmark | Claude Fable 5 | GPT-4o |
|---|---|---|
| AA Agentic Index | 56.6 | 9.7 |
| AA Intelligence | 62.1 | 11.2 |
| AA-LCR | 76.7 | 55 |
| AA-Omniscience | 43.3 | -10.5 |
| arc_agi_1 | 90.5 | 4.5 |
| ARC-AGI-2 | 76.8 | 0 |
| arena_vision | 1307 | 1162 |
| Artificial Analysis Coding Index | 76.5 | 24.2 |
| critpt | 28.6 | 0 |
| gdpval | 61.9 | 0 |
| GPQA Diamond | 92.6 | 70.1 |
| HLE | 64.5 | 5.3 |
| IFBench | 63.5 | 36 |
| mathvision | 94.8 | 30.4 |
| MMMU-Pro | 84.2 | 59.9 |
| OmniScience Accuracy | 65.3 | 23.7 |
| OmniScience Non-Hallucination | 36.4 | 62.1 |
| OSWorld-Verified | 85 | 5 |
| scicode | 60.2 | 33.4 |
| simplebench | 81.9 | 17.8 |
| SWE-bench Verified | 95 | 38.8 |
| Terminal-Bench Hard | 62.9 | 8.3 |
| τ²-Bench Telecom (AA run) | 98.5 | 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.