| Benchmark | Claude Sonnet 4.6 | Muse Spark 1.1 |
|---|---|---|
| AA Agentic Index | 61.6 | 39.7 |
| AA Intelligence | 48.4 | 53.2 |
| AA-LCR | 74 | 81.3 |
| AA-Omniscience | 12.2 | 28.1 |
| arena_elo | 1472 | 1491 |
| arena_text_factuality | 1460 | 1466 |
| Artificial Analysis Coding Index | 63 | 71.3 |
| charxiv_reasoning_with_tools | 85.3 | 88.4 |
| coding_arena_elo | 1523 | 1539 |
| critpt | 3.1 | 15.1 |
| DeepSWE 1.1 | 30 | 53.3 |
| Finance Agent v2 | 51 | 57.2 |
| gdpval | 54.8 | 43.7 |
| GPQA Diamond | 89.9 | 89.8 |
| healthbench_professional | 44.2 | 59.3 |
| HLE | 49 | 62.1 |
| livebench_agentic_coding | 42.6 | 58.5 |
| livebench_coding | 79.3 | 77.2 |
| livebench_data_analysis | 78 | 72.5 |
| livebench_instruction_following | 63.2 | 69.6 |
| livebench_language | 76.1 | 74.3 |
| livebench_math | 87 | 87.1 |
| livebench_reasoning | 84.8 | 87.7 |
| MCP Atlas | 69.5 | 88.1 |
| OmniScience Accuracy | 40.9 | 52 |
| OmniScience Non-Hallucination | 51.6 | 50 |
| OSWorld-Verified | 78.5 | 80.8 |
| scicode | 47 | 58.2 |
| SWE-bench Pro | 58.1 | 61.5 |
| TauBench V3 - Banking | 34.4 | 31.8 |
| Terminal-Bench 2.1 | 71.2 | 80 |
| toolathlon | 49.4 | 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.