| Benchmark | DeepSeek-V3.2 | Qwen3.7 Plus Preview |
|---|---|---|
| AA Agentic Index | 39.8 | 20.8 |
| AA Intelligence | 32.8 | 39.4 |
| AA-LCR | 70.7 | 69 |
| AA-Omniscience | -22.5 | 2.4 |
| apexAgents | 14.5 | 22.4 |
| arena_text_factuality | 1433 | 1454 |
| Artificial Analysis Coding Index | 44.2 | 55.9 |
| critpt | 2.9 | 9.1 |
| gdpval | 18.8 | 22.2 |
| GPQA Diamond | 84 | 90.3 |
| HLE | 40.8 | 35.6 |
| hmmt_feb_2026 | 84.1 | 92.9 |
| IFBench | 61 | 79.1 |
| imo_answer_bench | 78.3 | 86 |
| LiveCodeBench v6 | 83.3 | 89.6 |
| MCP Atlas | 62.2 | 73.2 |
| mcpmark | 38 | 58.7 |
| mmlu_redux | 93.7 | 94.5 |
| MMLU-Pro | 86 | 88.5 |
| OmniScience Accuracy | 33 | 22.5 |
| OmniScience Non-Hallucination | 17.3 | 74.5 |
| scicode | 39 | 51.3 |
| SWE-bench Multilingual | 70.2 | 75.8 |
| SWE-bench Pro | 15.6 | 57.6 |
| SWE-bench Verified | 73.1 | 77.7 |
| TauBench V3 - Banking | 18.8 | 17.9 |
| Terminal-Bench 2.0 | 46.4 | 70.3 |
| Terminal-Bench 2.1 | 46.8 | 61 |
| Terminal-Bench Hard | 35.6 | 47 |
| τ²-Bench Telecom (AA run) | 90.6 | 93 |
| τ³-Bench | 69.2 | 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.