| Benchmark | GPT-5.6 Sol | Muse Spark 1.1 |
|---|---|---|
| AA Agentic Index | 57.8 | 39.7 |
| AA Intelligence | 61 | 53.2 |
| AA-LCR | 77.7 | 81.3 |
| AA-Omniscience | 22 | 28.1 |
| arena_elo | 1482 | 1491 |
| arena_text_factuality | 1474 | 1466 |
| Artificial Analysis Coding Index | 78.3 | 71.3 |
| coding_arena_elo | 1619 | 1539 |
| critpt | 32.3 | 15.1 |
| cybergym | 83.6 | 59 |
| DeepSWE 1.1 | 73 | 53.3 |
| Finance Agent v2 | 53.8 | 57.2 |
| gdpval | 61.1 | 43.7 |
| GPQA Diamond | 94.6 | 89.8 |
| healthbench_professional | 60.5 | 59.3 |
| HLE | 49.5 | 62.1 |
| JobBench | 45.4 | 54.7 |
| livebench_agentic_coding | 56.2 | 58.5 |
| livebench_coding | 83.9 | 77.2 |
| livebench_data_analysis | 79.8 | 72.5 |
| livebench_instruction_following | 71.8 | 69.6 |
| livebench_language | 87.7 | 74.3 |
| livebench_math | 96.2 | 87.1 |
| livebench_reasoning | 91.7 | 87.7 |
| MCP Atlas | 83.6 | 88.1 |
| OmniScience Accuracy | 59.4 | 52 |
| OmniScience Non-Hallucination | 10.6 | 50 |
| OSWorld-Verified | 83 | 80.8 |
| scicode | 56.9 | 58.2 |
| SWE-bench Pro | 64.6 | 61.5 |
| TauBench V3 - Banking | 44.3 | 31.8 |
| Terminal-Bench 2.1 | 89.5 | 80 |
| toolathlon | 58 | 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.