3 ms·

However, the model card for GPT 5.2 looks amazing, wish I could actually see that performance in action!