it is based on the input prompt itself, not the number of output tokens

→ View original post on X — @officiallogank