| Benchmark | Kimi K3 | Muse Spark 1.1 |
|---|---|---|
| AA Agentic Index | 54.3 | 39.7 |
| AA Intelligence | 60 | 53.2 |
| AA-LCR | 82.7 | 81.3 |
| AA-Omniscience | 46 | 28.1 |
| arena_elo | 1489 | 1491 |
| arena_text_factuality | 1471 | 1466 |
| Artificial Analysis Coding Index | 76.2 | 71.3 |
| baby_vision | 85.7 | 76.3 |
| coding_arena_elo | 1674 | 1539 |
| critpt | 23.4 | 15.1 |
| cybergym | 80 | 59 |
| deepsearchqa_f1 | 95 | 84.9 |
| DeepSWE 1.1 | 69 | 53.3 |
| Finance Agent v2 | 54.4 | 57.2 |
| gdpval | 59.4 | 43.7 |
| GPQA Diamond | 93.5 | 89.8 |
| HLE | 56 | 62.1 |
| JobBench | 54.3 | 54.7 |
| livebench_agentic_coding | 62.2 | 58.5 |
| livebench_coding | 81.5 | 77.2 |
| livebench_data_analysis | 78.7 | 72.5 |
| livebench_instruction_following | 71.4 | 69.6 |
| livebench_language | 85.5 | 74.3 |
| livebench_math | 84.4 | 87.1 |
| livebench_reasoning | 90.7 | 87.7 |
| MCP Atlas | 84.2 | 88.1 |
| OmniScience Accuracy | 47.6 | 52 |
| OmniScience Non-Hallucination | 49.1 | 50 |
| OSWorld-Verified | 84.8 | 80.8 |
| scicode | 58.7 | 58.2 |
| TauBench V3 - Banking | 46 | 31.8 |
| Terminal-Bench 2.1 | 88.3 | 80 |
| toolathlon | 73.2 | 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.