| Benchmark | Claude Sonnet 4.6 | Qwen3.7 Max |
|---|---|---|
| AA Agentic Index | 61.6 | 30.9 |
| AA Intelligence | 48.4 | 46.7 |
| AA-LCR | 74 | 74.7 |
| AA-Omniscience | 12.2 | 13.5 |
| arena_elo | 1472 | 1474 |
| arena_text_factuality | 1460 | 1479 |
| Artificial Analysis Coding Index | 63 | 66 |
| coding_arena_elo | 1523 | 1517 |
| critpt | 3.1 | 13.4 |
| DeepSWE 1.1 | 30 | 21.6 |
| Finance Agent v2 | 51 | 48.4 |
| gdpval | 54.8 | 38.5 |
| GPQA Diamond | 89.9 | 92.4 |
| HLE | 49 | 41.4 |
| HLE (with tools) | 46.8 | 53.5 |
| IFBench | 56.6 | 80.5 |
| itbenchSre | 39.8 | 42.5 |
| livebench | 75.5 | 74.3 |
| livebench_agentic_coding | 42.6 | 43.6 |
| livebench_coding | 79.3 | 74.2 |
| livebench_data_analysis | 78 | 71.8 |
| livebench_instruction_following | 63.2 | 74 |
| livebench_language | 76.1 | 79.7 |
| livebench_math | 87 | 85.3 |
| livebench_reasoning | 84.8 | 83.3 |
| MCP Atlas | 69.5 | 76.4 |
| mmmlu | 89.3 | 90.3 |
| OmniScience Accuracy | 40.9 | 31.1 |
| OmniScience Non-Hallucination | 51.6 | 74.4 |
| scicode | 47 | 53.5 |
| simpleqa_verified | 29 | 58.5 |
| SWE-bench Pro | 58.1 | 60.6 |
| SWE-bench Verified | 79.6 | 80.4 |
| TauBench V3 - Banking | 34.4 | 11.8 |
| Terminal-Bench 2.0 | 59.1 | 69.7 |
| Terminal-Bench 2.1 | 71.2 | 75 |
| Terminal-Bench Hard | 59.1 | 50.8 |
| τ²-Bench Telecom (AA run) | 97.9 | 94.7 |
| τ³-Bench | 30.5 | 10.9 |
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.