The Assistants are more limited in comparison to custom GPTs, namely online search for example. I tried creating same kind of GPT with Assistant and Custom GPT, it’s not the same: so there’s no way of using an API to interact with a custom GPT?
I’d have thought it’s inevitable that at some point this will be implemented, and that you’ll be able to deploy a GPT to a website, or a slack channel, or a whatsapp number etc, through your OpenAI acount in a couple of clicks.
As @Foxabilo said, this is specifically what assistants are for? OpenAI likely won’t ever make the GPTs themselves interactable through the API, that defeats the whole purpose of the GPT store and the push for more Plus Subscribers.
Assistants are nearly identical, especially if you add in dalle-3 and code interpreter. Vision will likely come eventually, but at the moment it would be too compute-intensive to run on the scale of GPT-4 Turbo.
This is needed. i spent hours and hours yesterday creating a custom GPT client through the interface, So happy with my results only to find out that i have no way of connecting my fastapi script to it. i can only share the web link… what is the point ? hours and hours wasted.