For voice, it appears we will have two distinct models (Standard 4o and Advanced). It also seems the backend will determine which model to use based on your rate limit, access level, and other factors.

→ View original post on X — @testingcatalog