| Benchmark | Muse Spark | Opus 5 |
|---|---|---|
| AA Agentic Index | 28.7 | 59.2 |
| AA Intelligence | 44.3 | 63.1 |
| AA-LCR | 77 | 78.7 |
| AA-Omniscience | 7.2 | 37.1 |
| ARC-AGI-2 | 42.5 | 90.4 |
| arena_elo | 1488 | 1487 |
| arena_text_factuality | 1467 | 1481 |
| Artificial Analysis Coding Index | 58.6 | 78 |
| coding_arena_elo | 1540 | 1662 |
| critpt | 11.3 | 29.1 |
| gdpval | 32.3 | 67.2 |
| GPQA Diamond | 89.5 | 93.7 |
| HLE | 58.4 | 64.7 |
| MCP Atlas | 82.2 | 85.8 |
| MMMU-Pro | 80.5 | 84.7 |
| OmniScience Accuracy | 49.6 | 60.9 |
| OmniScience Non-Hallucination | 26.8 | 40.5 |
| scicode | 51.5 | 55.7 |
| simpleqa_verified | 66.3 | 56.7 |
| SWE-bench Pro | 55 | 79.2 |
| SWE-bench Verified | 77.4 | 96 |
| TauBench V3 - Banking | 19.6 | 44.7 |
| Terminal-Bench 2.1 | 62.2 | 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.