| Benchmark | Gemini 1.5 Pro | GPT-5.2 |
|---|---|---|
| AA Intelligence | 10 | 43.3 |
| ARC-AGI-2 | 0.8 | 52.9 |
| arena_vision | 1179 | 1245 |
| Artificial Analysis Coding Index | 23.6 | 48.7 |
| Fortress | 53.9 | 17.5 |
| GPQA Diamond | 59.1 | 92.4 |
| HLE | 4.9 | 45.5 |
| LiveCodeBench | 30.5 | 89 |
| MathVista | 70.6 | 82.8 |
| MMLU-Pro | 75.8 | 87 |
| MMMU-Pro | 55 | 79.5 |
| OCRBench | 800 | 80.7 |
| ocrbench_v2 | 51.6 | 50.5 |
| scicode | 29.5 | 52.1 |
| SEAL VISTA | 37.1 | 46.6 |
| simplebench | 27.1 | 45.8 |
| SWE-bench Verified | 34.2 | 80 |
| Video-MME | 78.6 | 86 |
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.