| Benchmark | Gemini 3.5 Flash | GPT-4o |
|---|---|---|
| AA Agentic Index | 70.4 | 9.7 |
| AA Intelligence | 52 | 11.2 |
| AA-LCR | 81 | 55 |
| AA-Omniscience | 21.2 | -10.5 |
| arc_agi_1 | 92.5 | 4.5 |
| ARC-AGI-2 | 72.1 | 0 |
| Artificial Analysis Coding Index | 70.1 | 24.2 |
| critpt | 13.1 | 0 |
| gdpval | 57.8 | 0 |
| GPQA Diamond | 92.2 | 70.1 |
| HLE | 42.7 | 5.3 |
| IFBench | 76.3 | 36 |
| MMMU-Pro | 84.3 | 59.9 |
| OmniScience Accuracy | 51.4 | 23.7 |
| OmniScience Non-Hallucination | 38.2 | 62.1 |
| OSWorld-Verified | 78.4 | 5 |
| scicode | 53.1 | 33.4 |
| simplebench | 76.7 | 17.8 |
| Terminal-Bench Hard | 46.2 | 8.3 |
| τ²-Bench Telecom (AA run) | 95.6 | 28.9 |
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.