| Benchmark | GPT-5.6 Sol | Qwen3.6 Plus |
|---|---|---|
| AA Agentic Index | 57.8 | 29 |
| AA Intelligence | 61 | 40.5 |
| AA-LCR | 77.7 | 72.3 |
| AA-Omniscience | 22 | 2.6 |
| aime_2026 | 99.9 | 95.3 |
| arena_elo | 1482 | 1444 |
| arena_text_factuality | 1474 | 1438 |
| Artificial Analysis Coding Index | 78.3 | 54.5 |
| coding_arena_elo | 1619 | 1460 |
| critpt | 32.3 | 2.9 |
| Finance Agent v2 | 53.8 | 40.9 |
| gdpval | 61.1 | 32 |
| GPQA Diamond | 94.6 | 90.4 |
| HLE | 49.5 | 28.8 |
| HLE (with tools) | 58 | 50.6 |
| IFBench | 72.7 | 75.2 |
| livebench_agentic_coding | 56.2 | 41.4 |
| livebench_coding | 83.9 | 78.2 |
| livebench_data_analysis | 79.8 | 69.9 |
| livebench_instruction_following | 71.8 | 58.3 |
| livebench_language | 87.7 | 75 |
| livebench_math | 96.2 | 83.7 |
| livebench_reasoning | 91.7 | 75.8 |
| longbench_v2 | 67.1 | 62 |
| mathvision | 95.8 | 88 |
| MCP Atlas | 83.6 | 74.1 |
| MMMU-Pro | 84.6 | 78.8 |
| OmniScience Accuracy | 59.4 | 26.4 |
| OmniScience Non-Hallucination | 10.6 | 68 |
| OSWorld-Verified | 83 | 62.5 |
| scicode | 56.9 | 40.7 |
| simpleqa_verified | 71.6 | 49.1 |
| SWE-bench Pro | 64.6 | 56.6 |
| TauBench V3 - Banking | 44.3 | 20.8 |
| Terminal-Bench 2.1 | 89.5 | 61.4 |
| Terminal-Bench Hard | 65.9 | 43.9 |
| toolathlon | 58 | 39.8 |
| Video-MME | 89.5 | 84.2 |
| τ²-Bench Telecom (AA run) | 85.1 | 97.7 |
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.