| Benchmark | Llama 3.1 Instruct 405B | Qwen3.6 27B |
|---|---|---|
| AA Agentic Index | 6.3 | 27.5 |
| AA Intelligence | 8.5 | 37.7 |
| AA-LCR | 24.7 | 73.3 |
| AA-Omniscience | -17.1 | -20 |
| Artificial Analysis Coding Index | 14.5 | 53.7 |
| C-Eval | 72.5 | 91.4 |
| critpt | 0 | 1.1 |
| gdpval | 0 | 32 |
| GPQA Diamond | 51.5 | 87.8 |
| hellaswag | 89.2 | 90 |
| HLE | 4.2 | 24 |
| humaneval | 89 | 78.2 |
| IFBench | 39 | 69.1 |
| mmlu_redux | 86.2 | 93.5 |
| MMLU-Pro | 73.4 | 86.2 |
| OmniScience Accuracy | 23.2 | 19.6 |
| OmniScience Non-Hallucination | 49 | 50.7 |
| scicode | 29.9 | 39.8 |
| simpleqa | 23.2 | 95.7 |
| SWE-bench Pro | 11.2 | 53.5 |
| SWE-bench Verified | 24.5 | 77.2 |
| Terminal-Bench Hard | 6.8 | 34.8 |
| τ²-Bench Telecom (AA run) | 19 | 94.2 |
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.