| Benchmark | Claude Opus 4.7 | Muse Spark 1.1 |
|---|---|---|
| AA Agentic Index | 64.6 | 39.7 |
| AA Intelligence | 55 | 53.2 |
| AA-LCR | 75.3 | 81.3 |
| AA-Omniscience | 27.3 | 28.1 |
| arena_elo | 1494 | 1491 |
| arena_text_factuality | 1474 | 1466 |
| Artificial Analysis Coding Index | 73.6 | 71.3 |
| charxiv_reasoning_with_tools | 91 | 88.4 |
| coding_arena_elo | 1557 | 1539 |
| critpt | 12 | 15.1 |
| cybergym | 73.8 | 59 |
| Finance Agent v2 | 51.5 | 57.2 |
| gdpval | 58.6 | 43.7 |
| GPQA Diamond | 94.2 | 89.8 |
| HLE | 54.7 | 62.1 |
| livebench_agentic_coding | 50.7 | 58.5 |
| livebench_coding | 82.1 | 77.2 |
| livebench_data_analysis | 78.3 | 72.5 |
| livebench_instruction_following | 66.7 | 69.6 |
| livebench_language | 77.9 | 74.3 |
| livebench_math | 92.8 | 87.1 |
| livebench_reasoning | 87.2 | 87.7 |
| MCP Atlas | 79.1 | 88.1 |
| OmniScience Accuracy | 48.9 | 52 |
| OmniScience Non-Hallucination | 57.7 | 50 |
| OSWorld-Verified | 82.8 | 80.8 |
| scicode | 54.5 | 58.2 |
| SWE-bench Pro | 64.3 | 61.5 |
| TauBench V3 - Banking | 34.6 | 31.8 |
| Terminal-Bench 2.1 | 83.1 | 80 |
| toolathlon | 59.3 | 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.