| Benchmark | Claude 4 Opus | Claude Fable 5 |
|---|---|---|
| AA Intelligence | 31.7 | 62.1 |
| AA-LCR | 40 | 76.7 |
| arc_agi_1 | 27 | 90.5 |
| ARC-AGI-2 | 8.6 | 76.8 |
| arena_vision | 1207 | 1307 |
| Artificial Analysis Coding Index | 34 | 76.5 |
| GPQA Diamond | 79.6 | 92.6 |
| HLE | 12.3 | 64.5 |
| IFBench | 53.7 | 63.5 |
| livebench | 74.6 | 78.3 |
| scicode | 40.9 | 60.2 |
| simplebench | 58.8 | 81.9 |
| SWE-bench Verified | 79.4 | 95 |
| Terminal-Bench Hard | 31.1 | 62.9 |
| τ²-Bench Telecom (AA run) | 73.4 | 98.5 |
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.