| Benchmark | DeepSeek-V4-Pro | Qwen3.7 Plus Preview |
|---|---|---|
| AA Agentic Index | 63.3 | 20.8 |
| AA Intelligence | 53 | 39.4 |
| AA-LCR | 70 | 69 |
| AA-Omniscience | -10.6 | 2.4 |
| Agents' Last Exam | 16.5 | 33.6 |
| Apex (Pass@1) | 38.3 | 22.7 |
| apexAgents | 24.3 | 22.4 |
| arena_elo | 1457 | 1458 |
| arena_text_factuality | 1450 | 1454 |
| Artificial Analysis Coding Index | 59.4 | 55.9 |
| critpt | 13 | 9.1 |
| gdpval | 49 | 22.2 |
| GPQA Diamond | 90.5 | 90.3 |
| HLE | 48.2 | 35.6 |
| hmmt_feb_2026 | 95.2 | 92.9 |
| IFBench | 76.5 | 79.1 |
| imo_answer_bench | 89.8 | 86 |
| MCP Atlas | 74.2 | 73.2 |
| MMLU-Pro | 87.5 | 88.5 |
| nl2repo | 38.5 | 41.1 |
| OmniScience Accuracy | 43 | 22.5 |
| OmniScience Non-Hallucination | 12.2 | 74.5 |
| scicode | 50 | 51.3 |
| SWE-bench Multilingual | 76.2 | 75.8 |
| SWE-bench Pro | 55.4 | 57.6 |
| SWE-bench Verified | 80.6 | 77.7 |
| TauBench V3 - Banking | 30.1 | 17.9 |
| Terminal-Bench 2.0 | 67.9 | 70.3 |
| Terminal-Bench 2.1 | 72.1 | 61 |
| Terminal-Bench Hard | 46.2 | 47 |
| τ²-Bench Telecom (AA run) | 96.2 | 93 |
| τ³-Bench | 25.8 | 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.