| Benchmark | Claude Sonnet 4.6 | Gemini 3.6 Flash |
|---|---|---|
| AA Agentic Index | 61.6 | 40.5 |
| AA Intelligence | 48.4 | 52 |
| AA-LCR | 74 | 79 |
| AA-Omniscience | 12.2 | 22.1 |
| arc_agi_1 | 86.5 | 34.5 |
| ARC-AGI-2 | 60.4 | 2.6 |
| arena_elo | 1472 | 1481 |
| arena_text_factuality | 1460 | 1470 |
| Artificial Analysis Coding Index | 63 | 69.2 |
| charxiv_reasoning_with_tools | 85.3 | 89.4 |
| charxiv_rq | 72.4 | 85.2 |
| coding_arena_elo | 1523 | 1538 |
| critpt | 3.1 | 10.6 |
| DeepSWE 1.1 | 30 | 49 |
| GDP (Surge AI) | 78.6 | 22 |
| gdpval | 54.8 | 46.1 |
| GDPval-AA v2 | 1381 | 1421 |
| GPQA Diamond | 89.9 | 92.8 |
| HLE | 49 | 40.8 |
| livebench_agentic_coding | 42.6 | 43.4 |
| livebench_coding | 79.3 | 77.9 |
| livebench_data_analysis | 78 | 63 |
| livebench_instruction_following | 63.2 | 75.4 |
| livebench_language | 76.1 | 83.9 |
| livebench_math | 87 | 86.4 |
| livebench_reasoning | 84.8 | 85.2 |
| MMMU-Pro | 75.6 | 83.2 |
| mrcr_v2_8needle_128k_average | 84.9 | 91.8 |
| OmniScience Accuracy | 40.9 | 50 |
| OmniScience Non-Hallucination | 51.6 | 44.4 |
| OSWorld-Verified | 78.5 | 83 |
| scicode | 47 | 52.7 |
| simpleqa_verified | 29 | 68.7 |
| SWE-bench Pro | 58.1 | 58.7 |
| TauBench V3 - Banking | 34.4 | 29.9 |
| Terminal-Bench 2.1 | 71.2 | 78 |
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.