Quick question for anyone who's tuned model choice on Vapi before — I've got an AI receptionist running on GPT-4.1 (latency ~1500ms). I tried swapping to GPT-4.1 mini (~1410ms) and GPT-4.1 nano (~1260ms) to cut latency. My concern is that the assistant follows a fairly complex prompt (pricing logic, safety rules, specific canned responses for different scenarios), and I don't want to lose accuracy for the sake of speed. Has anyone compared mini/nano vs full 4.1 for something with this level of prompt complexity? Is the accuracy drop noticeable in practice, or is it usually fine for receptionist-style use cases? Or should I skip both and use a different model entirely?

Hey everyone! I'm Zain, based in Birmingham UK. I run an AI receptionist service called Never Miss The Call. I've just built an AI receptionist called Lexi using Vapi for a sports therapy clinic — she's fully live and taking calls through Vonage. The one thing I'm stuck on is the Make.com SMS automation. When a call ends, I want Make.com to send the business owner a clean summary SMS with the caller's name, service requested, preferred day and time, and contact number. Has anyone done this before or does anyone have any video suggestions that show how to set this up? Any help would be massively appreciated! 🙏