| Benchmark | Gemini 3.5 Flash | GPT-5.6 Sol |
|---|---|---|
| AA Agentic Index | 70.4 | 57.8 |
| AA Intelligence | 52 | 61 |
| AA-LCR | 81 | 77.7 |
| AA-Omniscience | 21.2 | 22 |
| aime_2026 | 95 | 99.9 |
| arc_agi_1 | 92.5 | 97.5 |
| ARC-AGI-2 | 72.1 | 92.5 |
| arena_elo | 1477 | 1482 |
| arena_text_factuality | 1466 | 1474 |
| Artificial Analysis Coding Index | 70.1 | 78.3 |
| charxiv_rq | 84.2 | 87.8 |
| coding_arena_elo | 1490 | 1619 |
| critpt | 13.1 | 32.3 |
| DeepSWE | 37 | 73 |
| DeepSWE 1.1 | 37 | 73 |
| Finance Agent v2 | 57.9 | 53.8 |
| gdpval | 57.8 | 61.1 |
| GDPval-AA v2 | 1357 | 1748 |
| GPQA Diamond | 92.2 | 94.6 |
| HiL-Bench | 27.7 | 32.3 |
| HLE | 42.7 | 49.5 |
| IFBench | 76.3 | 72.7 |
| itbenchSre | 40.3 | 56.2 |
| livebench_agentic_coding | 49 | 56.2 |
| livebench_coding | 78.2 | 83.9 |
| livebench_data_analysis | 64.9 | 79.8 |
| livebench_instruction_following | 75.6 | 71.8 |
| livebench_language | 84.6 | 87.7 |
| livebench_math | 88.2 | 96.2 |
| livebench_reasoning | 82 | 91.7 |
| MCP Atlas | 83.6 | 83.6 |
| MMMU-Pro | 84.3 | 84.6 |
| OmniScience Accuracy | 51.4 | 59.4 |
| OmniScience Non-Hallucination | 38.2 | 10.6 |
| OSWorld-Verified | 78.4 | 83 |
| scicode | 53.1 | 56.9 |
| simplebench | 76.7 | 64.8 |
| simpleqa_verified | 68.4 | 71.6 |
| SWE-bench Pro | 55.1 | 64.6 |
| TauBench V3 - Banking | 32.2 | 44.3 |
| Terminal-Bench 2.1 | 78.7 | 89.5 |
| Terminal-Bench Hard | 46.2 | 65.9 |
| toolathlon | 56.5 | 58 |
| τ²-Bench Telecom (AA run) | 95.6 | 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.