| Benchmark | Gemini 3.6 Flash | GPT-5.1 |
|---|---|---|
| AA Agentic Index | 40.5 | 32.2 |
| AA Intelligence | 52 | 37.5 |
| AA-LCR | 79 | 76.7 |
| AA-Omniscience | 22.1 | 5.4 |
| arc_agi_1 | 34.5 | 72.8 |
| ARC-AGI-2 | 2.6 | 17.6 |
| arena_elo | 1481 | 1439 |
| arena_text_factuality | 1470 | 1445 |
| Artificial Analysis Coding Index | 69.2 | 49.4 |
| coding_arena_elo | 1538 | 1421 |
| critpt | 10.6 | 4.9 |
| gdpval | 46.1 | 25 |
| GPQA Diamond | 92.8 | 88.1 |
| HLE | 40.8 | 28.5 |
| matharena_visual_math_overall | 87.8 | 76.9 |
| MMMU-Pro | 83.2 | 75.5 |
| OmniScience Accuracy | 50 | 37.7 |
| OmniScience Non-Hallucination | 44.4 | 48.1 |
| scicode | 52.7 | 43.3 |
| simpleqa_verified | 68.7 | 48.9 |
| TauBench V3 - Banking | 29.9 | 15.9 |
| Terminal-Bench 2.1 | 78 | 52.4 |
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.