| Benchmark | Gemini 3 Flash Preview | Muse Spark |
|---|---|---|
| AA Agentic Index | 49.7 | 28.7 |
| AA Intelligence | 38.7 | 44.3 |
| AA-LCR | 73 | 77 |
| AA-Omniscience | 10.1 | 7.2 |
| ARC-AGI-2 | 33.6 | 42.5 |
| arena_elo | 1473 | 1488 |
| arena_text_factuality | 1452 | 1467 |
| arena_vision | 1260 | 1295 |
| Artificial Analysis Coding Index | 42.6 | 58.6 |
| coding_arena_elo | 1387 | 1540 |
| critpt | 8.6 | 11.3 |
| frontiermath_tier_4 | 4.2 | 14.6 |
| gdpval | 35.2 | 32.3 |
| GPQA Diamond | 90.4 | 89.5 |
| HLE | 43.5 | 58.4 |
| IFBench | 78 | 75.9 |
| MCP Atlas | 62 | 82.2 |
| MMMU-Pro | 81.2 | 80.5 |
| OmniScience Accuracy | 53.4 | 49.6 |
| OmniScience Non-Hallucination | 7.6 | 26.8 |
| scicode | 50.6 | 51.5 |
| screenspot_pro_no_tools | 69.1 | 84.1 |
| simpleqa_verified | 67.4 | 66.3 |
| SWE-bench Pro | 49.6 | 55 |
| SWE-bench Verified | 78 | 77.4 |
| TauBench V3 - Banking | 20.8 | 19.6 |
| Terminal-Bench 2.0 | 47.6 | 59 |
| Terminal-Bench 2.1 | 58 | 62.2 |
| Terminal-Bench Hard | 38.6 | 45.5 |
| τ²-Bench Telecom (AA run) | 80.4 | 91.5 |
| τ³-Bench | 17.5 | 19.6 |
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.