[jmorgan]

This is true! Although I'm also really excited at the potential speed (both for loading the model and token generation) of a 1B model for things like code completion.