Does anybody actually know how much unique AI slop you can create with one model?

So, I am running an experiment to test to see how much unique AI slop a model can actually produce and obviously it's producing quite a bit. Does anybody have any clue at the "mathematical theory of AI slop production?" How much slop should I expect from a 685gb model? Anybody have any clue how many petabytes of storage I'm going to need for this task or is it just going to produce like a "quasar of AI slop?" Where, it's technically going to just keep producing more and more variation as I do things like adjust the temperature? I'm kind of guessing that is what is going to occur, but obviously there has to be a limit... edit: Just text slop.