Interesting! How many token/s using what model are you getting out of it? And what are the other specs?