| Benchmark | Claude Opus 4.5 | Llama 3.1 Instruct 405B |
|---|---|---|
| AA Agentic Index | 59.6 | 6.3 |
| AA Intelligence | 41.9 | 8.5 |
| AA-LCR | 76 | 24.7 |
| AA-Omniscience | 14 | -17.1 |
| Artificial Analysis Coding Index | 47.8 | 14.5 |
| C-Eval | 92.2 | 72.5 |
| critpt | 4.6 | 0 |
| Fortress | 13.6 | 20.6 |
| gdpval | 47.3 | 0 |
| GPQA Diamond | 87 | 51.5 |
| HLE | 30.8 | 4.2 |
| IFBench | 58 | 39 |
| LiveCodeBench | 87 | 30.1 |
| longbench_v2 | 64.4 | 36.1 |
| mmlu_redux | 95.6 | 86.2 |
| MMLU-Pro | 90 | 73.4 |
| mmmlu | 90.8 | 73.8 |
| OmniScience Accuracy | 46.6 | 23.2 |
| OmniScience Non-Hallucination | 39 | 49 |
| scicode | 50 | 29.9 |
| simplebench | 62 | 23 |
| SWE-bench Pro | 57.1 | 11.2 |
| SWE-bench Verified | 81.5 | 24.5 |
| Terminal-Bench Hard | 47 | 6.8 |
| τ²-Bench Telecom (AA run) | 89.5 | 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.