Custom GPT: “Content failed to load” when switching models

Tested it with a custom GPT from my set-up, and both thinking and instantdo call my own (remote) API server, 'though.