| Benchmark | Claude Opus 4.7 | Llama 3.1 Instruct 405B |
|---|---|---|
| AA Agentic Index | 64.6 | 6.3 |
| AA Intelligence | 55 | 8.5 |
| AA-LCR | 75.3 | 24.7 |
| AA-Omniscience | 27.3 | -17.1 |
| Artificial Analysis Coding Index | 73.6 | 14.5 |
| critpt | 12 | 0 |
| gdpval | 58.6 | 0 |
| GPQA Diamond | 94.2 | 51.5 |
| harmbench | 17.5 | 62.7 |
| HLE | 54.7 | 4.2 |
| IFBench | 58.6 | 39 |
| mmmlu | 91.5 | 73.8 |
| OmniScience Accuracy | 48.9 | 23.2 |
| OmniScience Non-Hallucination | 57.7 | 49 |
| scicode | 54.5 | 29.9 |
| simplebench | 61.7 | 23 |
| SWE-bench Pro | 64.3 | 11.2 |
| SWE-bench Verified | 87.6 | 24.5 |
| Terminal-Bench Hard | 54.5 | 6.8 |
| τ²-Bench Telecom (AA run) | 88.6 | 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.