| Benchmark | Kimi K3 | Opus 5 |
|---|---|---|
| AA Agentic Index | 54.3 | 59.2 |
| AA Intelligence | 60 | 63.1 |
| AA-LCR | 82.7 | 78.7 |
| AA-Omniscience | 46 | 37.1 |
| arc_agi_1 | 65.7 | 97.5 |
| ARC-AGI-2 | 12.4 | 90.4 |
| arena_elo | 1489 | 1487 |
| arena_text_factuality | 1471 | 1481 |
| Artificial Analysis Coding Index | 76.2 | 78 |
| browsecomp | 91.2 | 90.8 |
| coding_arena_elo | 1674 | 1662 |
| critpt | 23.4 | 29.1 |
| DeepSWE 1.1 | 69 | 68.8 |
| Finance Agent v2 | 54.4 | 58.6 |
| gdpval | 59.4 | 67.2 |
| GDPval-AA v2 | 1687 | 1861 |
| GPQA Diamond | 93.5 | 93.7 |
| HLE | 56 | 64.7 |
| livebench_agentic_coding | 62.2 | 65.2 |
| livebench_coding | 81.5 | 81.5 |
| livebench_data_analysis | 78.7 | 74.5 |
| livebench_instruction_following | 71.4 | 63.8 |
| livebench_language | 85.5 | 88.7 |
| livebench_math | 84.4 | 95.7 |
| livebench_reasoning | 90.7 | 91.2 |
| MCP Atlas | 84.2 | 85.8 |
| MMMU-Pro | 83.4 | 84.7 |
| officeqa_pro | 63.3 | 66.9 |
| OmniScience Accuracy | 47.6 | 60.9 |
| OmniScience Non-Hallucination | 49.1 | 40.5 |
| OSWorld 2.0 | 58.3 | 70.6 |
| OSWorld-Verified | 84.8 | 4 |
| scicode | 58.7 | 55.7 |
| simplebench | 60.7 | 80.6 |
| simpleqa_verified | 42.7 | 56.7 |
| TauBench V3 - Banking | 46 | 44.7 |
| Terminal-Bench 2.1 | 88.3 | 89.1 |
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.