I believe I read 1 token is about half an English word. That is a few bytes at most. So what is the rest of a token memory actually used for?   submitted by   /u/Fancy-Snow7 [link]   [comments]