| Benchmark | GPT-5.4 | o4-mini |
|---|---|---|
| AA Agentic Index | 58.2 | 36.1 |
| AA Intelligence | 53.1 | 26.1 |
| AA-LCR | 77.7 | 60 |
| AA-Omniscience | 5.8 | -35.7 |
| arc_agi_1 | 93.7 | 58.7 |
| ARC-AGI-2 | 74 | 6.1 |
| arena_vision | 1277 | 1201 |
| Artificial Analysis Coding Index | 71.1 | 25.6 |
| browsecomp | 82.7 | 51.5 |
| critpt | 23.4 | 0.6 |
| frontiermath_tier_4 | 27.1 | 6.3 |
| gdpval | 50.1 | 25.4 |
| GPQA Diamond | 93 | 81.4 |
| HLE | 43.7 | 16.5 |
| IFBench | 73.9 | 68.7 |
| MMMU-Pro | 81.2 | 69.2 |
| MultiNRC | 58.3 | 22.2 |
| OmniScience Accuracy | 50.9 | 24.8 |
| OmniScience Non-Hallucination | 17.4 | 19.5 |
| scicode | 56.6 | 46.5 |
| SEAL VISTA | 50.9 | 51.8 |
| simpleqa_verified | 45.3 | 23.9 |
| Terminal-Bench Hard | 57.6 | 15.2 |
| vectara_answer_rate | 99.9 | 99.2 |
| vectara_avg_summary_length | 81.7 | 130.9 |
| vectara_factual_consistency | 93 | 81.4 |
| vectara_hallucination_rate ↓ | 7 | 18.6 |
| VisualToolBench | 29.2 | 11.1 |
| τ²-Bench Telecom (AA run) | 87.1 | 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.