| Benchmark | Claude Opus 4.6 | Llama 3.1 Instruct 405B |
|---|---|---|
| AA Agentic Index | 67.6 | 6.3 |
| AA Intelligence | 44.9 | 8.5 |
| AA-LCR | 74.3 | 24.7 |
| AA-Omniscience | 13.7 | -17.1 |
| Artificial Analysis Coding Index | 48.1 | 14.5 |
| Chinese SimpleQA (C-SimpleQA) | 76.4 | 54.7 |
| critpt | 12.6 | 0 |
| Fortress | 20.5 | 20.6 |
| gdpval | 55.9 | 0 |
| GPQA Diamond | 91.3 | 51.5 |
| GSM8K | 0.6 | 96.8 |
| HLE | 53.1 | 4.2 |
| IFBench | 62.5 | 39 |
| LiveCodeBench | 88.8 | 30.1 |
| MMLU-Pro | 89.1 | 73.4 |
| mmmlu | 91.1 | 73.8 |
| OmniScience Accuracy | 47 | 23.2 |
| OmniScience Non-Hallucination | 37.2 | 49 |
| scicode | 52 | 29.9 |
| simplebench | 67.6 | 23 |
| SWE-bench Pro | 57.3 | 11.2 |
| SWE-bench Verified | 80.8 | 24.5 |
| Terminal-Bench Hard | 65.4 | 6.8 |
| τ²-Bench Telecom (AA run) | 99.3 | 19 |
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.