3 ms·

I'll definitely believe that for video generation models, but those are also very compute-intensive for rather middling results.