| Benchmark | Claude Fable 5 | DeepSeek-V3 |
|---|---|---|
| AA Agentic Index | 56.6 | 1.6 |
| AA Intelligence | 62.1 | 15.4 |
| AA-LCR | 76.7 | 41.3 |
| AA-Omniscience | 43.3 | -37.6 |
| Artificial Analysis Coding Index | 76.5 | 23 |
| critpt | 28.6 | 0 |
| gdpval | 61.9 | 0 |
| GPQA Diamond | 92.6 | 68.4 |
| HLE | 64.5 | 5.2 |
| IFBench | 63.5 | 41 |
| livebench | 78.3 | 72.4 |
| OmniScience Accuracy | 65.3 | 25.4 |
| OmniScience Non-Hallucination | 36.4 | 23.3 |
| scicode | 60.2 | 35.8 |
| simplebench | 81.9 | 27.2 |
| SWE-bench Verified | 95 | 42 |
| TauBench V3 - Banking | 38.1 | 4.7 |
| Terminal-Bench 2.1 | 88 | 16.9 |
| Terminal-Bench Hard | 62.9 | 15.2 |
| τ²-Bench Telecom (AA run) | 98.5 | 47.1 |
| τ³-Bench | 26.8 | 4.7 |
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.