| Benchmark | Claude Opus 4.7 | Qwen3.6 27B |
|---|---|---|
| AA Agentic Index | 64.6 | 27.5 |
| AA Intelligence | 55 | 37.7 |
| AA-LCR | 75.3 | 73.3 |
| AA-Omniscience | 27.3 | -20 |
| aime_2026 | 95.8 | 94.1 |
| Artificial Analysis Coding Index | 73.6 | 53.7 |
| charxiv_rq | 82.1 | 78.4 |
| critpt | 12 | 1.1 |
| gdpval | 58.6 | 32 |
| GPQA Diamond | 94.2 | 87.8 |
| HLE | 54.7 | 24 |
| hmmt_feb_2026 | 93.9 | 84.3 |
| IFBench | 58.6 | 69.1 |
| livebench_agentic_coding | 50.7 | 39.3 |
| livebench_coding | 82.1 | 71.8 |
| livebench_data_analysis | 78.3 | 70.4 |
| livebench_instruction_following | 66.7 | 53.2 |
| livebench_language | 77.9 | 63.3 |
| livebench_math | 92.8 | 79.9 |
| livebench_reasoning | 87.2 | 70.3 |
| MMMU-Pro | 78.8 | 75.8 |
| OmniScience Accuracy | 48.9 | 19.6 |
| OmniScience Non-Hallucination | 57.7 | 50.7 |
| OSWorld-Verified | 82.8 | 63.9 |
| scicode | 54.5 | 39.8 |
| SWE-bench Multilingual | 80.5 | 71.3 |
| SWE-bench Pro | 64.3 | 53.5 |
| SWE-bench Verified | 87.6 | 77.2 |
| TauBench V3 - Banking | 34.6 | 16.7 |
| Terminal-Bench 2.0 | 69.4 | 59.3 |
| Terminal-Bench 2.1 | 83.1 | 60.7 |
| Terminal-Bench Hard | 54.5 | 34.8 |
| τ²-Bench Telecom (AA run) | 88.6 | 94.2 |
| τ³-Bench | 28.9 | 15.3 |
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.