| Benchmark | Gemini 3 Flash Preview | Muse Spark 1.1 |
|---|---|---|
| AA Agentic Index | 49.7 | 39.7 |
| AA Intelligence | 38.7 | 53.2 |
| AA-LCR | 73 | 81.3 |
| AA-Omniscience | 10.1 | 28.1 |
| arena_elo | 1473 | 1491 |
| arena_text_factuality | 1452 | 1466 |
| Artificial Analysis Coding Index | 42.6 | 71.3 |
| coding_arena_elo | 1387 | 1539 |
| critpt | 8.6 | 15.1 |
| Finance Agent v2 | 42.6 | 57.2 |
| gdpval | 35.2 | 43.7 |
| GPQA Diamond | 90.4 | 89.8 |
| HLE | 43.5 | 62.1 |
| MCP Atlas | 62 | 88.1 |
| OmniScience Accuracy | 53.4 | 52 |
| OmniScience Non-Hallucination | 7.6 | 50 |
| OSWorld-Verified | 65.1 | 80.8 |
| scicode | 50.6 | 58.2 |
| SWE-bench Pro | 49.6 | 61.5 |
| TauBench V3 - Banking | 20.8 | 31.8 |
| Terminal-Bench 2.1 | 58 | 80 |
| toolathlon | 49.4 | 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.