| Benchmark | Kimi K3 | Muse Spark |
|---|---|---|
| AA Agentic Index | 54.3 | 28.7 |
| AA Intelligence | 60 | 44.3 |
| AA-LCR | 82.7 | 77 |
| AA-Omniscience | 46 | 7.2 |
| ARC-AGI-2 | 12.4 | 42.5 |
| arena_elo | 1489 | 1488 |
| arena_text_factuality | 1471 | 1467 |
| Artificial Analysis Coding Index | 76.2 | 58.6 |
| coding_arena_elo | 1674 | 1540 |
| critpt | 23.4 | 11.3 |
| cybergym | 80 | 43.5 |
| deepsearchqa_f1 | 95 | 74.8 |
| gdpval | 59.4 | 32.3 |
| GPQA Diamond | 93.5 | 89.5 |
| HLE | 56 | 58.4 |
| itbenchSre | 47.7 | 16.2 |
| MCP Atlas | 84.2 | 82.2 |
| MMMU-Pro | 83.4 | 80.5 |
| OmniScience Accuracy | 47.6 | 49.6 |
| OmniScience Non-Hallucination | 49.1 | 26.8 |
| scicode | 58.7 | 51.5 |
| simpleqa_verified | 42.7 | 66.3 |
| TauBench V3 - Banking | 46 | 19.6 |
| Terminal-Bench 2.1 | 88.3 | 62.2 |
| ZeroBench | 0.4 | 0.3 |
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.