| Benchmark | Gemini 3.6 Flash | GPT-5 |
|---|---|---|
| AA Agentic Index | 40.5 | 49.7 |
| AA Intelligence | 52 | 35.3 |
| AA-LCR | 79 | 76.3 |
| AA-Omniscience | 22.1 | -8.7 |
| arc_agi_1 | 34.5 | 56.2 |
| ARC-AGI-2 | 2.6 | 7.5 |
| Artificial Analysis Coding Index | 69.2 | 38.9 |
| coding_arena_elo | 1538 | 1419 |
| critpt | 10.6 | 5.7 |
| gdpval | 46.1 | 32.5 |
| GPQA Diamond | 92.8 | 87.3 |
| HLE | 40.8 | 28.5 |
| matharena_visual_math_overall | 87.8 | 78.8 |
| MMMU-Pro | 83.2 | 78.4 |
| OmniScience Accuracy | 50 | 40.3 |
| OmniScience Non-Hallucination | 44.4 | 21 |
| scicode | 52.7 | 43 |
| simpleqa_verified | 68.7 | 50.6 |
| SWE-bench Pro | 58.7 | 41.8 |
| TauBench V3 - Banking | 29.9 | 22.1 |
| Terminal-Bench 2.1 | 78 | 35.2 |
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.