| Benchmark | Gemini 3.6 Flash | Muse Spark 1.1 |
|---|---|---|
| AA Agentic Index | 40.5 | 39.7 |
| AA Intelligence | 52 | 53.2 |
| AA-LCR | 79 | 81.3 |
| AA-Omniscience | 22.1 | 28.1 |
| arena_elo | 1481 | 1491 |
| arena_text_factuality | 1470 | 1466 |
| Artificial Analysis Coding Index | 69.2 | 71.3 |
| charxiv_reasoning_with_tools | 89.4 | 88.4 |
| coding_arena_elo | 1538 | 1539 |
| critpt | 10.6 | 15.1 |
| DeepSWE 1.1 | 49 | 53.3 |
| gdpval | 46.1 | 43.7 |
| GPQA Diamond | 92.8 | 89.8 |
| HLE | 40.8 | 62.1 |
| livebench_agentic_coding | 43.4 | 58.5 |
| livebench_coding | 77.9 | 77.2 |
| livebench_data_analysis | 63 | 72.5 |
| livebench_instruction_following | 75.4 | 69.6 |
| livebench_language | 83.9 | 74.3 |
| livebench_math | 86.4 | 87.1 |
| livebench_reasoning | 85.2 | 87.7 |
| OmniScience Accuracy | 50 | 52 |
| OmniScience Non-Hallucination | 44.4 | 50 |
| OSWorld-Verified | 83 | 80.8 |
| scicode | 52.7 | 58.2 |
| SWE-bench Pro | 58.7 | 61.5 |
| TauBench V3 - Banking | 29.9 | 31.8 |
| Terminal-Bench 2.1 | 78 | 80 |
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.