Thanks for the answers. One more for you. When the KB answers, that isn’t GPT3.5 answering - it’s way less verbose with a different tone. What model does the KB use?