I am testing the Actions feature for GPTs to write a review, but I keep hitting the usage cap for GPT-4. It is impossible to use this feature seriously with the cap in place.
You’ve reached the current usage cap for GPT-4, please try again after…
How can I switch the GPT to GPT-3.5?
Unfortunately there is no option to switch the model used by GPTs.
The biggest reason for this is that only GPT-4 has vision capabilities, Code interpreter, Dalle Image Generation, and web browsing.
My suggestion would be to either look at upgrading to a ChatGPT Team account (reportedly 100 requests per 3 hours instead of 40) or look for a way to test/troubleshoot your action without using the GPT directly.
It is rather frustrating, a couple minutes of testing can completely gobble up you prompts!
Thanks for the reply! In my view, this limits the usability of the Action feature dramatically. It appears to me that this hasn’t really been thought through.
I saw a huge potential in developing Custom GPTs with actions for professional enterprise applications, the range is infinite. OAuth feature makes it great.
After three months testing and testing, I still see the potential, but the big amount of issues plus the cap limitations makes me think to give up and abandon the projects, at least until limits are higher, if they will.
I understand technical reasons, this and that, and not even arguing, if there is a technical reason, there is. But then that makes impossible to continue.
Current cap limitations, even upgrading to the teams plan, makes this nonoperational for business purposes. Don’t forget that team caps is higher, but then there is a team… so same thing.
I’m going to pause all my projects for Custom GPTs so far, and wait to see if this gets better in the future.
There is no technical reason. If some features require GPT-4, you simple disable them when the user selects GTP-3.5 as model.
Is it possible to do Custom GPT’s with Actions calling to an external API using GPT-3.5 instead 4? with no cap limitations? if so, can you please let me know how? Thanks!
This method would not be feasible unfortunately since the actions are there for the GPT to call, so you would have to send a message to the GPT, thus still counting to the usage cap.
If you’re interested in using an API though you could use the Assistants API, they operate in a way similar to GPTs but without the web searching and image generation built-in.
Does Assistants API require a OpenAI Plus account or can be used with a free account and just pay per usage?
It’s completely separate from ChatGPT or any subscriptions, just sign up and pay per use.