| Benchmark | DeepSeek-V4-Pro | Qwen3.6 35B A3B |
|---|---|---|
| AA Agentic Index | 63.3 | 21.6 |
| AA Intelligence | 53 | 32.1 |
| AA-LCR | 70 | 66.7 |
| AA-Omniscience | -10.6 | -22.2 |
| aime_2026 | 96.7 | 92.7 |
| Artificial Analysis Coding Index | 59.4 | 41.9 |
| browsecomp | 83.4 | 48.7 |
| critpt | 13 | 0.3 |
| gdpval | 49 | 27.8 |
| GDPval-AA v2 | 1307 | 1015 |
| GPQA Diamond | 90.5 | 86 |
| HLE | 48.2 | 22.2 |
| hmmt_feb_2026 | 95.2 | 83.6 |
| hmmt_nov_2025 | 94.4 | 89.1 |
| IFBench | 76.5 | 64.4 |
| imo_answer_bench | 89.8 | 78.9 |
| MCP Atlas | 74.2 | 62.8 |
| MMLU-Pro | 87.5 | 85.6 |
| nl2repo | 38.5 | 29.4 |
| OmniScience Accuracy | 43 | 18.8 |
| OmniScience Non-Hallucination | 12.2 | 49.5 |
| scicode | 50 | 35.8 |
| SWE-bench Multilingual | 76.2 | 67.2 |
| SWE-bench Pro | 55.4 | 49.5 |
| SWE-bench Verified | 80.6 | 73.4 |
| TauBench V3 - Banking | 30.1 | 9.3 |
| Terminal-Bench 2.0 | 67.9 | 51.5 |
| Terminal-Bench 2.1 | 72.1 | 44.9 |
| Terminal-Bench Hard | 46.2 | 34.8 |
| tool_decathlon | 52.8 | 26.9 |
| toolathlon | 51.8 | 26.9 |
| τ²-Bench Telecom (AA run) | 96.2 | 95.3 |
| τ³-Bench | 25.8 | 67.2 |
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.