| Benchmark | Claude Sonnet 4.5 | Muse Spark 1.1 |
|---|---|---|
| AA Agentic Index | 50.6 | 39.7 |
| AA Intelligence | 37.4 | 53.2 |
| AA-LCR | 68.3 | 81.3 |
| AA-Omniscience | -0.1 | 28.1 |
| arena_elo | 1456 | 1491 |
| arena_text_factuality | 1448 | 1466 |
| Artificial Analysis Coding Index | 52.1 | 71.3 |
| coding_arena_elo | 1392 | 1539 |
| critpt | 1.1 | 15.1 |
| Fortress | 17 | 12.4 |
| gdpval | 40.4 | 43.7 |
| GPQA Diamond | 83.4 | 89.8 |
| HLE | 19.8 | 62.1 |
| MCP Atlas | 59.5 | 88.1 |
| multichallenge | 55.3 | 75.3 |
| MultiNRC | 35.8 | 65.6 |
| OmniScience Accuracy | 32.9 | 52 |
| OmniScience Non-Hallucination | 50.9 | 50 |
| OSWorld-Verified | 61.4 | 80.8 |
| scicode | 45 | 58.2 |
| SWE-bench Pro | 43.6 | 61.5 |
| TauBench V3 - Banking | 24.5 | 31.8 |
| Terminal-Bench 2.1 | 55.8 | 80 |
| toolathlon | 41 | 75.6 |
| VisualToolBench | 6.2 | 44.8 |
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.