| Benchmark | Gemini 3.1 Pro | Muse Spark 1.1 |
|---|---|---|
| AA Agentic Index | 23 | 39.7 |
| AA Intelligence | 47.7 | 53.2 |
| AA-LCR | 79 | 81.3 |
| AA-Omniscience | 32.9 | 28.1 |
| AIRS-Bench | 83 | 77 |
| arena_elo | 1486 | 1491 |
| arena_text_factuality | 1471 | 1466 |
| Artificial Analysis Coding Index | 68.8 | 71.3 |
| baby_vision | 51.6 | 76.3 |
| charxiv_reasoning_with_tools | 83.2 | 88.4 |
| coding_arena_elo | 1447 | 1539 |
| critpt | 17.7 | 15.1 |
| cybergym | 38.8 | 59 |
| deepsearchqa_f1 | 81.9 | 84.9 |
| DeepSWE 1.1 | 12 | 53.3 |
| Finance Agent v2 | 43 | 57.2 |
| Fortress | 29.8 | 12.4 |
| gdpval | 23.3 | 43.7 |
| GPQA Diamond | 94.3 | 89.8 |
| HLE | 51.4 | 62.1 |
| HLE Calibration | 50.4 | 23.4 |
| livebench_agentic_coding | 44.1 | 58.5 |
| livebench_coding | 76.5 | 77.2 |
| livebench_data_analysis | 78.5 | 72.5 |
| livebench_instruction_following | 79.1 | 69.6 |
| livebench_language | 85.4 | 74.3 |
| livebench_math | 91 | 87.1 |
| livebench_reasoning | 84 | 87.7 |
| MCP Atlas | 78.2 | 88.1 |
| multichallenge | 71.4 | 75.3 |
| MultiNRC | 64.7 | 65.6 |
| OmniScience Accuracy | 55.3 | 52 |
| OmniScience Non-Hallucination | 50.1 | 50 |
| OR-Bench (FRR) | 2.5 | 4.8 |
| OSWorld-Verified | 76.2 | 80.8 |
| scicode | 59 | 58.2 |
| SWE-bench Pro | 54.2 | 61.5 |
| TauBench V3 - Banking | 21.4 | 31.8 |
| Terminal-Bench 2.1 | 74 | 80 |
| toolathlon | 48.8 | 75.6 |
| VisualToolBench | 29 | 44.8 |
| wmdp | 86.3 | 87 |
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.