| Benchmark | GPT-5.5 | o4-mini |
|---|---|---|
| AA Agentic Index | 47.4 | 36.1 |
| AA Intelligence | 56.3 | 26.1 |
| AA-LCR | 79 | 60 |
| AA-Omniscience | 20.5 | -35.7 |
| arc_agi_1 | 95 | 58.7 |
| ARC-AGI-2 | 85 | 6.1 |
| arena_vision | 1283 | 1201 |
| Artificial Analysis Coding Index | 74.9 | 25.6 |
| browsecomp | 84.4 | 51.5 |
| critpt | 27.1 | 0.6 |
| Fortress | 16.3 | 21.5 |
| frontiermath_tier_4 | 35.4 | 6.3 |
| gdpval | 49.5 | 25.4 |
| GPQA Diamond | 93.6 | 81.4 |
| HLE | 52.2 | 16.5 |
| IFBench | 75.9 | 68.7 |
| MMMU-Pro | 83.2 | 69.2 |
| OmniScience Accuracy | 58 | 24.8 |
| OmniScience Non-Hallucination | 12 | 19.5 |
| scicode | 56.1 | 46.5 |
| simplebench | 69 | 38.7 |
| simpleqa_verified | 63.1 | 23.9 |
| SWE-bench Verified | 80.6 | 68.1 |
| Terminal-Bench Hard | 60.6 | 15.2 |
| vectara_answer_rate | 100 | 99.2 |
| vectara_avg_summary_length | 129.6 | 130.9 |
| vectara_factual_consistency | 90.7 | 81.4 |
| vectara_hallucination_rate ↓ | 9.3 | 18.6 |
| τ²-Bench Telecom (AA run) | 93.9 | 55.6 |
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.