| Benchmark | Gemini 2.5 Flash | GPT-5.6 Sol |
|---|---|---|
| AA Agentic Index | 18.8 | 57.8 |
| AA Intelligence | 17.7 | 61 |
| AA-LCR | 65.7 | 77.7 |
| AA-Omniscience | -29.8 | 22 |
| arc_agi_1 | 33.3 | 97.5 |
| ARC-AGI-2 | 2.5 | 92.5 |
| arena_text_factuality | 1437 | 1474 |
| Artificial Analysis Coding Index | 22.2 | 78.3 |
| critpt | 1.4 | 32.3 |
| gdpval | 11.9 | 61.1 |
| Global-MMLU-Lite | 88.4 | 91.8 |
| GPQA Diamond | 82.8 | 94.6 |
| HLE | 12.1 | 49.5 |
| IFBench | 50.3 | 72.7 |
| MMMU-Pro | 69.1 | 84.6 |
| OmniScience Accuracy | 26.1 | 59.4 |
| OmniScience Non-Hallucination | 24.7 | 10.6 |
| scicode | 35.9 | 56.9 |
| simplebench | 41.2 | 64.8 |
| Terminal-Bench Hard | 13.6 | 65.9 |
| τ²-Bench Telecom (AA run) | 31.6 | 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.