| Benchmark | Kimi K2.5 | Opus 5 |
|---|---|---|
| AA Agentic Index | 52.8 | 59.2 |
| AA Intelligence | 36 | 63.1 |
| AA-LCR | 73 | 78.7 |
| AA-Omniscience | -7.3 | 37.1 |
| arc_agi_1 | 65.3 | 97.5 |
| ARC-AGI-2 | 11.8 | 90.4 |
| arena_elo | 1450 | 1487 |
| arena_text_factuality | 1445 | 1481 |
| Artificial Analysis Coding Index | 46.8 | 78 |
| browsecomp | 74.9 | 90.8 |
| coding_arena_elo | 1436 | 1662 |
| critpt | 3.1 | 29.1 |
| gdpval | 38.3 | 67.2 |
| GDPval-AA v2 | 1009 | 1861 |
| GPQA Diamond | 87.9 | 93.7 |
| HLE | 50.2 | 64.7 |
| HLE (with tools) | 51.8 | 64.7 |
| MCP Atlas | 64 | 85.8 |
| MMMU-Pro | 78.5 | 84.7 |
| OmniScience Accuracy | 35.2 | 60.9 |
| OmniScience Non-Hallucination | 50 | 40.5 |
| OSWorld-Verified | 63.3 | 4 |
| scicode | 49 | 55.7 |
| simplebench | 46.8 | 80.6 |
| simpleqa_verified | 36.9 | 56.7 |
| SWE-bench Multilingual | 73 | 89.5 |
| SWE-bench Pro | 53.8 | 79.2 |
| SWE-bench Verified | 76.8 | 96 |
| TauBench V3 - Banking | 14.2 | 44.7 |
| Terminal-Bench 2.1 | 45.7 | 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.