| Benchmark | GPT-5.5 | Llama 3.1 Instruct 405B |
|---|---|---|
| AA Agentic Index | 47.4 | 6.3 |
| AA Intelligence | 56.3 | 8.5 |
| AA-LCR | 79 | 24.7 |
| AA-Omniscience | 20.5 | -17.1 |
| Artificial Analysis Coding Index | 74.9 | 14.5 |
| critpt | 27.1 | 0 |
| Fortress | 16.3 | 20.6 |
| gdpval | 49.5 | 0 |
| GPQA Diamond | 93.6 | 51.5 |
| HLE | 52.2 | 4.2 |
| IFBench | 75.9 | 39 |
| OmniScience Accuracy | 58 | 23.2 |
| OmniScience Non-Hallucination | 12 | 49 |
| scicode | 56.1 | 29.9 |
| simplebench | 69 | 23 |
| SWE-bench Pro | 58.6 | 11.2 |
| SWE-bench Verified | 80.6 | 24.5 |
| Terminal-Bench Hard | 60.6 | 6.8 |
| τ²-Bench Telecom (AA run) | 93.9 | 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.