Cursor AI calls the wrong model when Open API URL is over-ridden

[banderberg (Barry)]

I am using v0 in Cursor AI by overriding the Open API URL and specifying my v0 API key and choosing GPT 4o as my model. This has been working great.

After I did some UI work, I switched the model to Gemini 2.5 Pro and continued working. Eventually I started getting an error:

Request failed with status code 403: {“success”:false,“error”:“Usage-based pricing must be enabled to access the v0 API: v0 by Vercel”}

I opened a new chat and switched to Sonnet 3.7 and I get the same error.

It appears that Cursor is still trying to talk to v0 even though I am not using GPT 4 as my model.

What’s worse is that now I’ve spent all my v0 credits by unknowingly doing a bunch of stuff I wouldn’t use v0 for.

[deanrie]

Hey, try disabling your API key, there might be a conflict.

[banderberg (Barry)]

If I do that then I can’t use v0. What sort of conflict?

[deanrie]

You can enable and disable your API key if needed. You can use the shortcut cmd/ctrl + shift + 0 for this.

All requests go through our backend. Considering the specifics of this API key, when you don’t add a specific model to the list of models and use the gpt-4o model, it’s likely that the API key intercepts all requests, causing a conflict.