If I add a custom model like MiniMax by overriding the OpenAI API key and base URL, will it affect Cursor’s native OpenAI GPT models?
Will I still be able to use the built-in models, or does this replace them entirely?
Hey, good question.
No, adding a custom model doesn’t replace the built-in models. All built-in models (GPT, Claude, Gemini, etc.) will still be available in the dropdown as usual. The custom model just gets added to the list next to them.
There’s one important nuance though. There’s a known issue right now. If you enable Override OpenAI Base URL and enter your API Key, those credentials are incorrectly applied to all models, including Cursor proprietary ones like Composer 2. That can break them.
Workaround: toggle the API Key/Base URL override off before using built-in models. Quick access to settings is Ctrl+Shift+0.
More details: Composer 2 broken when custom OpenAI API Key is enabled
The team is aware, but there’s no ETA yet. Let me know if you have any questions.
I tried, but its breaks inbuilt models.
Cant use them.
Instend give another section to bring any custom provider.
Provider: XYZ
API Mode [openai | Antrophic | google genai]
base_url
model_list = [list of models] ie thats would show up in ui
in UI show model as provider_name/model_name
Thats it. how difficult is it?
just ask devs to prompt it.
Yeah, got it. The toggle workaround is pretty bad UX, so it’s not a real solution. The issue where Override Base URL leaks to all models, including Composer 2, is known and already reported internally.
The idea of a separate Custom Providers section with provider name, API mode, base_url, a model list, and showing provider/model in the list is reasonable. It also matches the direction we’re looking at to separate BYOK keys from the built-in model credentials.
I can’t share a timeline for a fix or redesign yet. If I get an update, I’ll post it in this thread.