| Benchmark | Gemini 3.5 Flash | o4-mini |
|---|---|---|
| AA Agentic Index | 70.4 | 36.1 |
| AA Intelligence | 52 | 26.1 |
| AA-LCR | 81 | 60 |
| AA-Omniscience | 21.2 | -35.7 |
| arc_agi_1 | 92.5 | 58.7 |
| ARC-AGI-2 | 72.1 | 6.1 |
| Artificial Analysis Coding Index | 70.1 | 25.6 |
| critpt | 13.1 | 0.6 |
| frontiermath_tier_4 | 14.6 | 6.3 |
| gdpval | 57.8 | 25.4 |
| GPQA Diamond | 92.2 | 81.4 |
| HLE | 42.7 | 16.5 |
| IFBench | 76.3 | 68.7 |
| MMMU-Pro | 84.3 | 69.2 |
| OmniScience Accuracy | 51.4 | 24.8 |
| OmniScience Non-Hallucination | 38.2 | 19.5 |
| scicode | 53.1 | 46.5 |
| simplebench | 76.7 | 38.7 |
| simpleqa_verified | 68.4 | 23.9 |
| Terminal-Bench Hard | 46.2 | 15.2 |
| τ²-Bench Telecom (AA run) | 95.6 | 55.6 |
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.