Speed of generating text in models like GPT-4

Hello everyone!!!

I was using GPT-4 as always for daily tasks and one question come up in my mind.

What determines speed of generating text and is it possible to somehow set some rules for example number of tokens generated in one sec?

I do not know if everyone is familiar with speed reading. In short story-long, this is a technique which allows human read faster by showing for example only one word at the time in the center of the screen and highlighting the center letter.

I am using GPT-4 most of the time not only, because it is generating better responses, but honestly I am able to read it while generating which allows me to work faster, because it’s somehow imitating this speed reading technique.

Thanks for all responses.