| Benchmark | GPT-4o | GPT-5.5 |
|---|---|---|
| AA Agentic Index | 9.7 | 47.4 |
| AA Intelligence | 11.2 | 56.3 |
| AA-LCR | 55 | 79 |
| AA-Omniscience | -10.5 | 20.5 |
| arc_agi_1 | 4.5 | 95 |
| ARC-AGI-2 | 0 | 85 |
| arena_vision | 1162 | 1283 |
| Artificial Analysis Coding Index | 24.2 | 74.9 |
| critpt | 0 | 27.1 |
| Fortress | 47.2 | 16.3 |
| gdpval | 0 | 49.5 |
| GPQA Diamond | 70.1 | 93.6 |
| HLE | 5.3 | 52.2 |
| IFBench | 36 | 75.9 |
| MMMU-Pro | 59.9 | 83.2 |
| OmniScience Accuracy | 23.7 | 58 |
| OmniScience Non-Hallucination | 62.1 | 12 |
| OSWorld-Verified | 5 | 78.7 |
| scicode | 33.4 | 56.1 |
| simplebench | 17.8 | 69 |
| SWE-bench Verified | 38.8 | 80.6 |
| Terminal-Bench Hard | 8.3 | 60.6 |
| vectara_answer_rate | 93.8 | 100 |
| vectara_avg_summary_length | 86.6 | 129.6 |
| vectara_factual_consistency | 90.4 | 90.7 |
| vectara_hallucination_rate ↓ | 9.6 | 9.3 |
| τ²-Bench Telecom (AA run) | 28.9 | 93.9 |
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.