Cursor-managed models are routed through Override OpenAI Base URL

Support Bug Reports
,
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

When Override OpenAI Base URL is enabled, Cursor routes requests for Cursor-managed models through the configured custom OpenAI-compatible endpoint.

For example, when I select gpt-5.6-luna from Cursor’s built-in model picker, the request is still sent to my custom proxy instead of Cursor’s backend.

I confirmed this behavior through the proxy logs.

My intended setup is to use both Cursor subscription models and custom models from my own OpenAI-compatible proxy in the same Cursor profile.

Steps to Reproduce

  1. Configure a custom OpenAI-compatible Base URL in Cursor.
  2. Enable Override OpenAI Base URL.
  3. Select a Cursor-managed model from the built-in model picker, for example gpt-5.6-luna.
  4. Send a request.
  5. Check the custom proxy logs.
  6. The request for gpt-5.6-luna is sent to the custom proxy.

If gpt-5.6-luna is intentionally unavailable on the proxy, Cursor returns:

The requested model is not available

Expected Behavior

Cursor-managed models selected from the built-in model picker should continue to use Cursor’s backend.

Custom models should use the configured custom OpenAI Base URL.

Ideally, the routing should behave like this:

Cursor-managed model → Cursor backend

Custom model → Custom OpenAI Base URL

Enabling Override OpenAI Base URL should not redirect Cursor-managed subscription models to the custom endpoint.

Screenshots / Screen Recordings

image.png
image.png399×529 18.8 KB

Operating System

Linux

Version Information

Version: 3.12.17
VS Code Extension API: 1.128.0
Commit: 0fb762053c34788bb7760d5673f8a6d4c8589d50
Date: 2026-07-17T02:53:53.006Z
Layout: IDE
Build Type: Stable
Release Track: Default
Electron: 40.10.3
Chromium: 144.0.7559.236
Node.js: 24.15.0
V8: 14.4.258.32-electron.0
xterm.js: 6.1.0-beta.256
OS: Linux x64 6.17.0-40-generic

For AI issues: which model did you use?

GPT-5.6 Luna (gpt-5.6-luna)

For AI issues: add Request ID with privacy disabled

5e466056-a8a2-4b1e-b437-bb91234259fb

Additional Information

This behavior was also confirmed by Cursor staff in the original feature-request thread:

https://forum.cursor.com/t/custom-base-urls-for-each-custom-model/147219/13

Cursor staff explained that, with Override OpenAI Base URL enabled, all requests currently go through the custom endpoint, including Cursor-managed models from the built-in picker.

The current workaround is to manually disable Override OpenAI Base URL when using Cursor subscription models, and enable it again when using custom models.

I created this separate report at the request of Cursor staff so that this routing issue can be tracked independently from the original per-model Base URL feature request.

I can also provide proxy logs showing that requests for gpt-5.6-luna are reaching the custom endpoint.

Does this stop you from using Cursor

Sometimes - I can sometimes use Cursor

5

Hey, thanks for moving this into a separate thread and confirming it via proxy logs. That made it much easier for us to verify on our side.

What you’re seeing matches the current routing logic. While Override OpenAI Base URL is enabled, all OpenAI-family requests go through your custom endpoint, including models picked from the built-in picker like gpt-5.6-luna. Your config is correct. Right now, the feature works exactly like that, and there isn’t a carve-out for managed models yet.

We’re tracking the request for per-model routing, where built-in models would keep going to the Cursor backend and custom models would go to your endpoint. I can’t share a timeline yet, but I’ll reply in the thread once there’s an update.

For now, the workaround is the same. Turn off Override OpenAI Base URL when using Cursor models, then turn it back on for custom ones.