| Benchmark | Gemini 3 Flash Preview | GPT-5.6 Sol |
|---|---|---|
| AA Agentic Index | 49.7 | 57.8 |
| AA Intelligence | 38.7 | 61 |
| AA-LCR | 73 | 77.7 |
| AA-Omniscience | 10.1 | 22 |
| aime_2026 | 96.7 | 99.9 |
| arc_agi_1 | 84.7 | 97.5 |
| ARC-AGI-2 | 33.6 | 92.5 |
| arena_elo | 1473 | 1482 |
| arena_text_factuality | 1452 | 1474 |
| Artificial Analysis Coding Index | 42.6 | 78.3 |
| charxiv_rq | 80.3 | 87.8 |
| coding_arena_elo | 1387 | 1619 |
| critpt | 8.6 | 32.3 |
| Finance Agent v2 | 42.6 | 53.8 |
| gdpval | 35.2 | 61.1 |
| GPQA Diamond | 90.4 | 94.6 |
| HLE | 43.5 | 49.5 |
| IFBench | 78 | 72.7 |
| MCP Atlas | 62 | 83.6 |
| MMMU-Pro | 81.2 | 84.6 |
| OmniScience Accuracy | 53.4 | 59.4 |
| OmniScience Non-Hallucination | 7.6 | 10.6 |
| OSWorld-Verified | 65.1 | 83 |
| scicode | 50.6 | 56.9 |
| simplebench | 61.1 | 64.8 |
| simpleqa_verified | 67.4 | 71.6 |
| SWE-bench Pro | 49.6 | 64.6 |
| TauBench V3 - Banking | 20.8 | 44.3 |
| Terminal-Bench 2.1 | 58 | 89.5 |
| Terminal-Bench Hard | 38.6 | 65.9 |
| toolathlon | 49.4 | 58 |
| τ²-Bench Telecom (AA run) | 80.4 | 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.