| Benchmark | Gemini 2.5 Flash | Muse Spark |
|---|---|---|
| AA Agentic Index | 18.8 | 28.7 |
| AA Intelligence | 17.7 | 44.3 |
| AA-LCR | 65.7 | 77 |
| AA-Omniscience | -29.8 | 7.2 |
| ARC-AGI-2 | 2.5 | 42.5 |
| arena_text_factuality | 1437 | 1467 |
| arena_vision | 1214 | 1295 |
| Artificial Analysis Coding Index | 22.2 | 58.6 |
| critpt | 1.4 | 11.3 |
| frontiermath_tier_4 | 4.2 | 14.6 |
| gdpval | 11.9 | 32.3 |
| GPQA Diamond | 82.8 | 89.5 |
| HLE | 12.1 | 58.4 |
| IFBench | 50.3 | 75.9 |
| MMMU-Pro | 69.1 | 80.5 |
| OmniScience Accuracy | 26.1 | 49.6 |
| OmniScience Non-Hallucination | 24.7 | 26.8 |
| scicode | 35.9 | 51.5 |
| SWE-bench Verified | 60.4 | 77.4 |
| Terminal-Bench Hard | 13.6 | 45.5 |
| τ²-Bench Telecom (AA run) | 31.6 | 91.5 |
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.