Late-2025 GPT-5 refresh with improved reasoning and steerability.
Iterative improvement on GPT-5.1 with enhanced reasoning and faster performance
| Benchmark | GPT-5.2 | GPT-5.1 | Δ |
|---|---|---|---|
| MMLU | 92.8% | 92.5% | +0.3 |
| MATH | 88.5% | — | — |
| HumanEval | 95.8% | — | — |
| MMLU-Pro | 90.8% | 89.2% | +1.6 |
| GPQA Diamond | 80.1% | 77.8% | +2.3 |
| SWE-bench Verified | 72.5% | 70.1% | +2.4 |
| AIME 2025 | 92.1% | 100% | -7.9 |