Another issue with Store-GPTs in ChatGPT with GPT-5: In the mobile apps, calling actions breaks up after contacting endpoint. (It works in mobile browser and PC-app, so likely not an endpoint issue). The error message, after it looks like endpoint is contacted, is of a general kind that suggests the issue is with the endpoint. Any provided more to-the-point info from the instructions is ignored. This gives the impression, that the creator is at fault, also it provides not help how to work around the issue.
So, since GPT-5, store GPTs that rely on actions are broken in many different ways, with no way to fix it as the builder and with no clear info provided.
Voice mode transcription is mangled (only in GPTs, not in general chat)
Voice mode can’t reliably call actions
In the mobile apps action calling is disabled, with a forced general error that blames the endpoint
What is the objective with the store GPTs, if there are regressions on such a level, that what makes GPTs relevant in the first place: the provided data, the actions, the instructions, break or are ignored with a model update? While at the same time pulling alternative models, that would still work, completely and forcibly…
I am having issues with calling actions even with 4o now since Firday (specially calling POST requests). Prior to this I get a dialog asking me to allow the endpoint. But now dialog is missing and it hangs on the Talking to…