3 ms·

LLMs don't generate their language sequentially either, they just output it sequentially token by token.