| Benchmark | GPT-5.2 | Muse Spark 1.1 |
|---|---|---|
| AA Agentic Index | 60.2 | 39.7 |
| AA Intelligence | 43.3 | 53.2 |
| AA-LCR | 79.3 | 81.3 |
| AA-Omniscience | 0.3 | 28.1 |
| arena_elo | 1437 | 1491 |
| arena_text_factuality | 1442 | 1466 |
| Artificial Analysis Coding Index | 48.7 | 71.3 |
| coding_arena_elo | 1418 | 1539 |
| critpt | 11.6 | 15.1 |
| deepsearchqa_f1 | 71.3 | 84.9 |
| Fortress | 17.5 | 12.4 |
| gdpval | 48.3 | 43.7 |
| GPQA Diamond | 92.4 | 89.8 |
| HLE | 45.5 | 62.1 |
| livebench_agentic_coding | 50.3 | 58.5 |
| livebench_coding | 76.1 | 77.2 |
| livebench_data_analysis | 78.2 | 72.5 |
| livebench_instruction_following | 61.8 | 69.6 |
| livebench_language | 79.8 | 74.3 |
| livebench_math | 93.2 | 87.1 |
| livebench_reasoning | 83.2 | 87.7 |
| MCP Atlas | 68 | 88.1 |
| MultiNRC | 42.2 | 65.6 |
| OmniScience Accuracy | 44.3 | 52 |
| OmniScience Non-Hallucination | 38.4 | 50 |
| scicode | 52.1 | 58.2 |
| SWE-bench Pro | 55.6 | 61.5 |
| TauBench V3 - Banking | 11.1 | 31.8 |
| toolathlon | 46.3 | 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.