The same gateway, the same key;
gpt-4o-realtime-preview can be added;
gpt-5.4-mini / gpt-5.5 cannot be added;
Suspicion: The logic for adding custom models wrongly rejects the “.” in the ID.
Steps to Reproduce
The same gateway, the same key;
gpt-4o-realtime-preview can be added;
gpt-5.4-mini / gpt-5.5 cannot be added;
Suspicion: The logic for adding custom models wrongly rejects the “.” in the ID.
Hey, thanks for the report. The dot in the ID isn’t the cause. Client-side validation clearly allows dots like my.model, so it shouldn’t get stripped during adding.
It’s most likely one of two things. To figure out which, I need a bit more info:
A screenshot of the exact error message you see when trying to add gpt-5.4-mini or gpt-5.5.
Are you using a custom OpenAI Base URL plus your own key via a BYOK gateway, or are you just adding the model via Cursor’s standard access?
When you say “it won’t add”, does it fail while adding in Settings, or does it add but then not work when you try to use it?
Hypotheses:
gpt-5.4-mini and gpt-5.5 are already in Cursor’s built-in model catalog. If you try to add them as custom, you might get “The model is already available”. In that case, just enable them in Settings > Models instead of adding them as custom.
If this is BYOK via your own gateway with a model that’s not in our catalog, there’s a known backend validation issue with model names. No ETA for a fix yet.
One more thing. You’re on version 2.6.22 from late March. There have been a lot of updates to the model catalog and BYOK since then. Please update to the latest version via Help > Check for Updates and see if it still reproduces.