Has anyone ran into the issue where phone-based Vapi calls are defaulting to gpt-3.5-turbo-1106 even when the assistant is set to use gpt-4o or any other provider model. I've noticed this only happens on voice calls (phone calls) and not on web calls.