| Benchmark | Claude Opus 4.7 | Kimi K3 |
|---|---|---|
| AA Agentic Index | 64.6 | 54.3 |
| AA Intelligence | 55 | 60 |
| AA-LCR | 75.3 | 82.7 |
| AA-Omniscience | 27.3 | 46 |
| aime_2026 | 95.8 | 96.7 |
| arc_agi_1 | 92 | 65.7 |
| ARC-AGI-2 | 75.8 | 12.4 |
| arena_elo | 1494 | 1489 |
| arena_text_factuality | 1474 | 1471 |
| Artificial Analysis Coding Index | 73.6 | 76.2 |
| browsecomp | 79.8 | 91.2 |
| charxiv_rq | 82.1 | 84.8 |
| coding_arena_elo | 1557 | 1674 |
| critpt | 12 | 23.4 |
| cybergym | 73.8 | 80 |
| Finance Agent v2 | 51.5 | 54.4 |
| gdpval | 58.6 | 59.4 |
| GPQA Diamond | 94.2 | 93.5 |
| HLE | 54.7 | 56 |
| hmmt_feb_2026 | 93.9 | 97 |
| itbenchSre | 46.7 | 47.7 |
| livebench_agentic_coding | 50.7 | 62.2 |
| livebench_coding | 82.1 | 81.5 |
| livebench_data_analysis | 78.3 | 78.7 |
| livebench_instruction_following | 66.7 | 71.4 |
| livebench_language | 77.9 | 85.5 |
| livebench_math | 92.8 | 84.4 |
| livebench_reasoning | 87.2 | 90.7 |
| MCP Atlas | 79.1 | 84.2 |
| MMMU-Pro | 78.8 | 83.4 |
| officeqa_pro | 80.6 | 63.3 |
| OmniScience Accuracy | 48.9 | 47.6 |
| OmniScience Non-Hallucination | 57.7 | 49.1 |
| OSWorld-Verified | 82.8 | 84.8 |
| scicode | 54.5 | 58.7 |
| simplebench | 61.7 | 60.7 |
| simpleqa_verified | 50.6 | 42.7 |
| TauBench V3 - Banking | 34.6 | 46 |
| Terminal-Bench 2.1 | 83.1 | 88.3 |
| toolathlon | 59.3 | 73.2 |
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.