Cursur suddenly began sending all requests to OpenAI endpoint

Discussion
1

I’ve been using Cursor off and on for a while, using multiple APIs simultaneously.
For Deepseek, distilled versions of Deepseek, and Qwen, I’m using an OpenAI endpoint at DeepInfra.
For cursor-small and other autocomplete, as well as Gemini 2, I’ve been using my Cursor pro subscription.

Everything has been working fine, with auto-completions working great, cursor-small working, gemini 2 flash working in chat and composer, and when needed, I can easily switch to Deepseek or Qwen.

Suddenly, all my gemini, cursor, and autocompletions stopped and started giving errors. The error says “Unable to reach api.deepinfra.com” Even when trying to use cursor specific models, 01, Gemini, etc. The only models that work are The Deepinfra ones.

How do I get it back working the way it was? I’ve tried almost every combination of adding/removing/editing the models and api keys but no luck.

2

Hey, could this be a temporary issue? I suggest you contact your provider for assistance, they might have changed something on their end.

3

Unfortunately not, the problem seems to be that all requests are going to the OpenAI endpoint rather than Cursor’s:

image
image495×271 13.9 KB

This error just started today, and the past several days I was able to use models like Gemini 2 flash and Cursor small just fine alongside my custom models.

4

Try disabling all models and leave only those related to your API.