| Benchmark | Claude Opus 4.6 | Gemini 3.6 Flash |
|---|---|---|
| AA Agentic Index | 67.6 | 40.5 |
| AA Intelligence | 44.9 | 52 |
| AA-LCR | 74.3 | 79 |
| AA-Omniscience | 13.7 | 22.1 |
| aime_2026 | 96.7 | 96.7 |
| arc_agi_1 | 93 | 34.5 |
| ARC-AGI-2 | 68.8 | 2.6 |
| arena_elo | 1497 | 1481 |
| arena_text_factuality | 1487 | 1470 |
| Artificial Analysis Coding Index | 48.1 | 69.2 |
| charxiv_reasoning_with_tools | 84.7 | 89.4 |
| charxiv_rq | 69.1 | 85.2 |
| coding_arena_elo | 1545 | 1538 |
| critpt | 12.6 | 10.6 |
| gdpval | 55.9 | 46.1 |
| GPQA Diamond | 91.3 | 92.8 |
| HLE | 53.1 | 40.8 |
| hmmt_feb_2026 | 96.2 | 89.4 |
| livebench_agentic_coding | 49 | 43.4 |
| livebench_coding | 78.2 | 77.9 |
| livebench_data_analysis | 69.9 | 63 |
| livebench_instruction_following | 63.3 | 75.4 |
| livebench_language | 83.3 | 83.9 |
| livebench_math | 89.3 | 86.4 |
| livebench_reasoning | 88.7 | 85.2 |
| matharena_visual_math_overall | 72.3 | 87.8 |
| MMMU-Pro | 77.3 | 83.2 |
| mrcr_v2_8needle_128k_average | 84 | 91.8 |
| OmniScience Accuracy | 47 | 50 |
| OmniScience Non-Hallucination | 37.2 | 44.4 |
| OSWorld-Verified | 72.7 | 83 |
| scicode | 52 | 52.7 |
| simpleqa_verified | 46.5 | 68.7 |
| SWE-bench Pro | 57.3 | 58.7 |
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.