| Benchmark | Llama 3.1 Instruct 405B | Qwen3.7 Max |
|---|---|---|
| AA Agentic Index | 6.3 | 30.9 |
| AA Intelligence | 8.5 | 46.7 |
| AA-LCR | 24.7 | 74.7 |
| AA-Omniscience | -17.1 | 13.5 |
| Artificial Analysis Coding Index | 14.5 | 66 |
| critpt | 0 | 13.4 |
| gdpval | 0 | 38.5 |
| GPQA Diamond | 51.5 | 92.4 |
| HLE | 4.2 | 41.4 |
| IFBench | 39 | 80.5 |
| ifeval | 88.6 | 94.3 |
| longbench_v2 | 36.1 | 65.3 |
| mmlu_redux | 86.2 | 95 |
| MMLU-Pro | 73.4 | 89.6 |
| mmmlu | 73.8 | 90.3 |
| OmniScience Accuracy | 23.2 | 31.1 |
| OmniScience Non-Hallucination | 49 | 74.4 |
| scicode | 29.9 | 53.5 |
| simplebench | 23 | 70.4 |
| SWE-bench Pro | 11.2 | 60.6 |
| SWE-bench Verified | 24.5 | 80.4 |
| Terminal-Bench Hard | 6.8 | 50.8 |
| τ²-Bench Telecom (AA run) | 19 | 94.7 |
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.