3 ms·

I thought you could not compare tokens across models because their cost and speed was so different between models.