@BradWyatt, Ahh got it, that helps. Sounds like you’re hitting one of those broader “ChatGPT error/loading” edge cases.
What you’re seeing lines up with known troubleshooting patterns for site failures during actions like model switching. The common thread is usually something breaking in the session or blocked resources.
Quick recap of what’s worth trying:
If none of that helps, it’s probably deeper than a local setup issue. At that point, grabbing console logs + a HAR file with timestamps and sending it to support@openai.com is the best way to get it properly investigated.
-Mark G
Update:
I think I figured out what’s going on.
If I turn Actions OFF, everything works
If I turn Actions ON, the model selector crashes (“Content failed to load”)
This happens across multiple GPTs, not just one.
Support said Actions GPTs are supposed to only work with GPT-4o / GPT-4.1, but those don’t show up in the dropdown anymore — I only see GPT-5.x and o3.
So it seems like:
Actions expects older models
Those models aren’t selectable anymore
New models aren’t supported → UI breaks
Also seeing this in console:
TypeError: f.models.get is not a function
Separately, I’ve also been seeing inconsistent behavior with Actions actually executing (even before this), so wouldn’t be surprised if GPT-5.x just isn’t fully compatible with Actions yet.
For now, only workaround is turning Actions off.
Curious if anyone else is seeing the same thing.
@BradWyatt, yeah I get why it looks like a model support issue, especially with what you’re seeing in the UI and console.
The confusing part is it’s not really about “older vs newer models” being required. What’s actually happening is:
If the Action spec or auth config has any issues, that rebuild step can fail, which is why the UI feels like it’s breaking when switching models.
So it’s kind of two things overlapping:
That’s also why it works fine with Actions off.
If you want to double check, quick test:
If that works cleanly, then it confirms it’s the spec/auth causing the break, not the model restriction itself.
This one’s definitely not obvious from the UI, you’re not alone in hitting it.
-Mark G.
Hi. I confirmed the issue. Even adding anything in the action breaks the UI. Can you kindly inform if there Is any patch coming for it? Or is it going to stay like this? Currently any action in any gpt is unusable from my end. It is not authentication issue. But rather custom action issue. As @BradWyatt mentioned I am also getting the same error.
It would be great if you can inform us whether this problem will be addressed or not.
Thanks you.
Hi,
I confirmed the same issue with my custom gpts. Also with windows app.
I noticed this today when I wanted to use the latest published model.
Like other users, I have actions configured for each of the custom GPTs.
I tried to check my developers console as well, and when ‘trying to trigger the bug’ I mentioned that the bug was gone.
Didn’t change anything at my side, did not delete any cookie / cache, etc…
Even GPT’s with custom (remote) actions can now be switched by the pulldown-menu, without the bug / crash.
Oh, thank you for reply, OpenAI fixed it. I checked it
I can confirm that it looks like OpenAI fixed it.
However, I want to note that when I set the model to use Instant, it does not make API calls. The Thinking model does work for API calls though.
o3: Works
5.2 Instant: Does not work
5.2 Thinking: Works
Latest model Instant: Does not work
Latest model Thinking: Works
Can someone else in this thread please confirm that they’re seeing this behavior as well?
+, I have tested it too. Works fine with every model
