| Benchmark | DeepSeek-R1 | Qwen3.6 27B |
|---|---|---|
| AA Agentic Index | 3.1 | 27.5 |
| AA Intelligence | 18.6 | 37.7 |
| AA-LCR | 56 | 73.3 |
| AA-Omniscience | -31.3 | -20 |
| Artificial Analysis Coding Index | 24.6 | 53.7 |
| C-Eval | 91.8 | 91.4 |
| critpt | 0.6 | 1.1 |
| gdpval | 1.5 | 32 |
| GPQA Diamond | 81 | 87.8 |
| HLE | 17.7 | 24 |
| HMMT 2025 | 79.4 | 90.7 |
| hmmt_feb_2025 | 76.7 | 93.8 |
| IFBench | 39 | 69.1 |
| mmlu_redux | 93.4 | 93.5 |
| MMLU-Pro | 85 | 86.2 |
| OmniScience Accuracy | 30.7 | 19.6 |
| OmniScience Non-Hallucination | 10.5 | 50.7 |
| scicode | 35.7 | 39.8 |
| simpleqa | 92.3 | 95.7 |
| SWE-bench Multilingual | 30.5 | 71.3 |
| SWE-bench Verified | 57.6 | 77.2 |
| TauBench V3 - Banking | 6.4 | 16.7 |
| Terminal-Bench 2.1 | 19.1 | 60.7 |
| Terminal-Bench Hard | 6.1 | 34.8 |
| τ²-Bench Telecom (AA run) | 11.4 | 94.2 |
| τ³-Bench | 6.4 | 15.3 |
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.