| Benchmark | Gemini 2.5 Pro | GPT-5.6 Sol |
|---|---|---|
| AA Agentic Index | 7.2 | 57.8 |
| AA Intelligence | 27 | 61 |
| AA-LCR | 66 | 77.7 |
| AA-Omniscience | -14.3 | 22 |
| arc_agi_1 | 41 | 97.5 |
| ARC-AGI-2 | 4.9 | 92.5 |
| arena_elo | 1446 | 1482 |
| arena_text_factuality | 1462 | 1474 |
| Artificial Analysis Coding Index | 46.7 | 78.3 |
| browsecomp | 9.9 | 90.4 |
| coding_arena_elo | 1224 | 1619 |
| critpt | 2.6 | 32.3 |
| gdpval | 8.5 | 61.1 |
| Global-MMLU-Lite | 88.6 | 91.8 |
| GPQA Diamond | 84.4 | 94.6 |
| HLE | 22.5 | 49.5 |
| HLE (with tools) | 28.4 | 58 |
| IFBench | 49 | 72.7 |
| mathvision | 73.3 | 95.8 |
| MMMU-Pro | 74.9 | 84.6 |
| OmniScience Accuracy | 39 | 59.4 |
| OmniScience Non-Hallucination | 12.6 | 10.6 |
| scicode | 43 | 56.9 |
| simplebench | 62.4 | 64.8 |
| simpleqa_verified | 56 | 71.6 |
| TauBench V3 - Banking | 9.7 | 44.3 |
| Terminal-Bench 2.1 | 28.5 | 89.5 |
| Terminal-Bench Hard | 26.5 | 65.9 |
| Video-MME | 84.8 | 89.5 |
| τ²-Bench Telecom (AA run) | 54.1 | 85.1 |
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.