| Benchmark | Muse Spark 1.1 | Opus 5 |
|---|---|---|
| AA Agentic Index | 39.7 | 59.2 |
| AA Intelligence | 53.2 | 63.1 |
| AA-LCR | 81.3 | 78.7 |
| AA-Omniscience | 28.1 | 37.1 |
| arena_elo | 1491 | 1487 |
| arena_text_factuality | 1466 | 1481 |
| Artificial Analysis Coding Index | 71.3 | 78 |
| coding_arena_elo | 1539 | 1662 |
| critpt | 15.1 | 29.1 |
| DeepSWE 1.1 | 53.3 | 68.8 |
| Finance Agent v2 | 57.2 | 58.6 |
| gdpval | 43.7 | 67.2 |
| GPQA Diamond | 89.8 | 93.7 |
| healthbench_professional | 59.3 | 59.8 |
| HLE | 62.1 | 64.7 |
| livebench_agentic_coding | 58.5 | 65.2 |
| livebench_coding | 77.2 | 81.5 |
| livebench_data_analysis | 72.5 | 74.5 |
| livebench_instruction_following | 69.6 | 63.8 |
| livebench_language | 74.3 | 88.7 |
| livebench_math | 87.1 | 95.7 |
| livebench_reasoning | 87.7 | 91.2 |
| MCP Atlas | 88.1 | 85.8 |
| OmniScience Accuracy | 52 | 60.9 |
| OmniScience Non-Hallucination | 50 | 40.5 |
| OSWorld-Verified | 80.8 | 4 |
| scicode | 58.2 | 55.7 |
| SWE-bench Pro | 61.5 | 79.2 |
| TauBench V3 - Banking | 31.8 | 44.7 |
| Terminal-Bench 2.1 | 80 | 89.1 |
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.