| Benchmark | Kimi K2.5 | Muse Spark 1.1 |
|---|---|---|
| AA Agentic Index | 52.8 | 39.7 |
| AA Intelligence | 36 | 53.2 |
| AA-LCR | 73 | 81.3 |
| AA-Omniscience | -7.3 | 28.1 |
| arena_elo | 1450 | 1491 |
| arena_text_factuality | 1445 | 1466 |
| Artificial Analysis Coding Index | 46.8 | 71.3 |
| baby_vision | 36.5 | 76.3 |
| coding_arena_elo | 1436 | 1539 |
| critpt | 3.1 | 15.1 |
| cybergym | 41.3 | 59 |
| deepsearchqa_f1 | 89 | 84.9 |
| Fortress | 41.1 | 12.4 |
| gdpval | 38.3 | 43.7 |
| GPQA Diamond | 87.9 | 89.8 |
| HLE | 50.2 | 62.1 |
| MCP Atlas | 64 | 88.1 |
| multichallenge | 61.4 | 75.3 |
| MultiNRC | 35.2 | 65.6 |
| OmniScience Accuracy | 35.2 | 52 |
| OmniScience Non-Hallucination | 50 | 50 |
| OSWorld-Verified | 63.3 | 80.8 |
| scicode | 49 | 58.2 |
| SWE-bench Pro | 53.8 | 61.5 |
| TauBench V3 - Banking | 14.2 | 31.8 |
| Terminal-Bench 2.1 | 45.7 | 80 |
| toolathlon | 27.8 | 75.6 |
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.