| Benchmark | Claude Sonnet 4.6 | Qwen3.7 Plus Preview |
|---|---|---|
| AA Agentic Index | 61.6 | 20.8 |
| AA Intelligence | 48.4 | 39.4 |
| AA-LCR | 74 | 69 |
| AA-Omniscience | 12.2 | 2.4 |
| apexAgents | 28 | 22.4 |
| arena_elo | 1472 | 1458 |
| arena_text_factuality | 1460 | 1454 |
| arena_vision | 1278 | 1263 |
| Artificial Analysis Coding Index | 63 | 55.9 |
| charxiv_rq | 72.4 | 85.9 |
| critpt | 3.1 | 9.1 |
| DeepSWE 1.1 | 30 | 14.2 |
| Finance Agent v2 | 51 | 38.2 |
| gdpval | 54.8 | 22.2 |
| GPQA Diamond | 89.9 | 90.3 |
| HLE | 49 | 35.6 |
| IFBench | 56.6 | 79.1 |
| MCP Atlas | 69.5 | 73.2 |
| mmmlu | 89.3 | 89 |
| MMMU-Pro | 75.6 | 80.5 |
| OmniScience Accuracy | 40.9 | 22.5 |
| OmniScience Non-Hallucination | 51.6 | 74.5 |
| OSWorld-Verified | 78.5 | 73.3 |
| scicode | 47 | 51.3 |
| SWE-bench Pro | 58.1 | 57.6 |
| SWE-bench Verified | 79.6 | 77.7 |
| TauBench V3 - Banking | 34.4 | 17.9 |
| Terminal-Bench 2.0 | 59.1 | 70.3 |
| Terminal-Bench 2.1 | 71.2 | 61 |
| Terminal-Bench Hard | 59.1 | 47 |
| τ²-Bench Telecom (AA run) | 97.9 | 93 |
| τ³-Bench | 30.5 | 17.9 |
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.