| Benchmark | GPT-5.6 Sol | Muse Spark |
|---|---|---|
| AA Agentic Index | 57.8 | 28.7 |
| AA Intelligence | 61 | 44.3 |
| AA-LCR | 77.7 | 77 |
| AA-Omniscience | 22 | 7.2 |
| ARC-AGI-2 | 92.5 | 42.5 |
| arena_elo | 1482 | 1488 |
| arena_text_factuality | 1474 | 1467 |
| Artificial Analysis Coding Index | 78.3 | 58.6 |
| coding_arena_elo | 1619 | 1540 |
| critpt | 32.3 | 11.3 |
| cybergym | 83.6 | 43.5 |
| gdpval | 61.1 | 32.3 |
| GPQA Diamond | 94.6 | 89.5 |
| HLE | 49.5 | 58.4 |
| IFBench | 72.7 | 75.9 |
| itbenchSre | 56.2 | 16.2 |
| MCP Atlas | 83.6 | 82.2 |
| MMMU-Pro | 84.6 | 80.5 |
| OmniScience Accuracy | 59.4 | 49.6 |
| OmniScience Non-Hallucination | 10.6 | 26.8 |
| scicode | 56.9 | 51.5 |
| simpleqa_verified | 71.6 | 66.3 |
| SWE-bench Pro | 64.6 | 55 |
| TauBench V3 - Banking | 44.3 | 19.6 |
| Terminal-Bench 2.1 | 89.5 | 62.2 |
| Terminal-Bench Hard | 65.9 | 45.5 |
| τ²-Bench Telecom (AA run) | 85.1 | 91.5 |
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.