| Benchmark | Claude Opus 4.7 | Gemini 3.6 Flash |
|---|---|---|
| AA Agentic Index | 64.6 | 40.5 |
| AA Intelligence | 55 | 52 |
| AA-LCR | 75.3 | 79 |
| AA-Omniscience | 27.3 | 22.1 |
| aime_2026 | 95.8 | 96.7 |
| arc_agi_1 | 92 | 34.5 |
| ARC-AGI-2 | 75.8 | 2.6 |
| arena_elo | 1494 | 1481 |
| arena_text_factuality | 1474 | 1470 |
| Artificial Analysis Coding Index | 73.6 | 69.2 |
| charxiv_reasoning_with_tools | 91 | 89.4 |
| charxiv_rq | 82.1 | 85.2 |
| coding_arena_elo | 1557 | 1538 |
| critpt | 12 | 10.6 |
| gdpval | 58.6 | 46.1 |
| GPQA Diamond | 94.2 | 92.8 |
| HLE | 54.7 | 40.8 |
| hmmt_feb_2026 | 93.9 | 89.4 |
| livebench_agentic_coding | 50.7 | 43.4 |
| livebench_coding | 82.1 | 77.9 |
| livebench_data_analysis | 78.3 | 63 |
| livebench_instruction_following | 66.7 | 75.4 |
| livebench_language | 77.9 | 83.9 |
| livebench_math | 92.8 | 86.4 |
| livebench_reasoning | 87.2 | 85.2 |
| MMMU-Pro | 78.8 | 83.2 |
| mrcr_v2_8needle_128k_average | 59.3 | 91.8 |
| OmniScience Accuracy | 48.9 | 50 |
| OmniScience Non-Hallucination | 57.7 | 44.4 |
| OSWorld-Verified | 82.8 | 83 |
| scicode | 54.5 | 52.7 |
| simpleqa_verified | 50.6 | 68.7 |
| SWE-bench Pro | 64.3 | 58.7 |
| TauBench V3 - Banking | 34.6 | 29.9 |
| Terminal-Bench 2.1 | 83.1 | 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.