| Benchmark | Claude Fable 5 | Gemini 3.6 Flash |
|---|---|---|
| AA Agentic Index | 56.6 | 40.5 |
| AA Intelligence | 62.1 | 52 |
| AA-LCR | 76.7 | 79 |
| AA-Omniscience | 43.3 | 22.1 |
| arc_agi_1 | 90.5 | 34.5 |
| ARC-AGI-2 | 76.8 | 2.6 |
| arena_elo | 1508 | 1481 |
| arena_text_factuality | 1486 | 1470 |
| Artificial Analysis Coding Index | 76.5 | 69.2 |
| charxiv_rq | 89.4 | 85.2 |
| coding_arena_elo | 1626 | 1538 |
| critpt | 28.6 | 10.6 |
| DeepSWE | 70 | 49 |
| DeepSWE 1.1 | 70 | 49 |
| GDP (Surge AI) | 29.8 | 22 |
| gdpval | 61.9 | 46.1 |
| GDPval-AA v2 | 1760 | 1421 |
| GPQA Diamond | 92.6 | 92.8 |
| HLE | 64.5 | 40.8 |
| livebench_agentic_coding | 62.2 | 43.4 |
| livebench_coding | 86 | 77.9 |
| livebench_data_analysis | 80.5 | 63 |
| livebench_instruction_following | 75.8 | 75.4 |
| livebench_language | 90.7 | 83.9 |
| livebench_math | 96 | 86.4 |
| livebench_reasoning | 89.7 | 85.2 |
| MMMU-Pro | 84.2 | 83.2 |
| OmniScience Accuracy | 65.3 | 50 |
| OmniScience Non-Hallucination | 36.4 | 44.4 |
| OSWorld 2.0 | 66.1 | 33.8 |
| OSWorld-Verified | 85 | 83 |
| scicode | 60.2 | 52.7 |
| simpleqa_verified | 68.3 | 68.7 |
| SWE-bench Pro | 80 | 58.7 |
| TauBench V3 - Banking | 38.1 | 29.9 |
| Terminal-Bench 2.1 | 88 | 78 |
| Terminal-bench 3.0 | 34.1 | 5.4 |
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.