| Benchmark | o3 | Qwen3.6 Plus |
|---|---|---|
| AA Agentic Index | 36.1 | 29 |
| AA Intelligence | 31.1 | 40.5 |
| AA-LCR | 73.3 | 72.3 |
| AA-Omniscience | -15.3 | 2.6 |
| Artificial Analysis Coding Index | 38.4 | 54.5 |
| critpt | 1.1 | 2.9 |
| ERQA | 64 | 65.7 |
| frontiermath_tier_4 | 2.1 | 8.3 |
| gdpval | 12.8 | 32 |
| GPQA Diamond | 83.3 | 90.4 |
| HLE | 20.3 | 28.8 |
| IFBench | 71.4 | 75.2 |
| longbench_v2 | 58.8 | 62 |
| MMLU-Pro | 85 | 88.5 |
| MMMU | 82.9 | 86 |
| MMMU-Pro | 76.4 | 78.8 |
| OmniScience Accuracy | 38.6 | 26.4 |
| OmniScience Non-Hallucination | 12.9 | 68 |
| scicode | 41 | 40.7 |
| simpleqa_verified | 53 | 49.1 |
| SWE-bench Verified | 69.1 | 78.8 |
| Terminal-Bench Hard | 37.1 | 43.9 |
| VideoMMMU | 83.3 | 84 |
| τ²-Bench Telecom (AA run) | 80.7 | 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.