| Benchmark | Opus 5 | Qwen3.7 Plus Preview |
|---|---|---|
| AA Agentic Index | 59.2 | 20.8 |
| AA Intelligence | 63.1 | 39.4 |
| AA-LCR | 78.7 | 69 |
| AA-Omniscience | 37.1 | 2.4 |
| arena_elo | 1487 | 1458 |
| arena_text_factuality | 1481 | 1454 |
| Artificial Analysis Coding Index | 78 | 55.9 |
| critpt | 29.1 | 9.1 |
| DeepSWE 1.1 | 68.8 | 14.2 |
| Finance Agent v2 | 58.6 | 38.2 |
| gdpval | 67.2 | 22.2 |
| GPQA Diamond | 93.7 | 90.3 |
| HLE | 64.7 | 35.6 |
| MCP Atlas | 85.8 | 73.2 |
| MMMU-Pro | 84.7 | 80.5 |
| OmniScience Accuracy | 60.9 | 22.5 |
| OmniScience Non-Hallucination | 40.5 | 74.5 |
| OSWorld-Verified | 4 | 73.3 |
| scicode | 55.7 | 51.3 |
| SWE-bench Multilingual | 89.5 | 75.8 |
| SWE-bench Pro | 79.2 | 57.6 |
| SWE-bench Verified | 96 | 77.7 |
| TauBench V3 - Banking | 44.7 | 17.9 |
| Terminal-Bench 2.1 | 89.1 | 61 |
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.