Getting "model does not work" regularly while it's half way in Agent mode

[gemyagouatc (Eugene Myasyshchev)]

Describe the Bug

I’m using a custom OpenAI url and API key. It’s been working well until few days back it started failing half way with the message “model xxx doesn’t work with your current plan” or something.

It’s happening after it did few steps, so it starts the task, updates one or two files, and then hangs with this message.

When typing resume - it goes on, but then fails again.

Steps to Reproduce

Use custom OpenAI url with custom model. Start some task in Agent mode that is relatively big and updates few files.

Expected Behavior

It iterates on the task without disruptions

Screenshots / Screen Recordings

Operating System

MacOS

Current Cursor Version (Menu → About Cursor → Copy)

Version: 1.3.7 (Universal)
VSCode Version: 1.99.3
Commit: 5d56108ff2401c28f239aa2a527ef424ce7719a0
Date: 2025-07-31T04:04:12.156Z (1 day ago)
Electron: 34.5.1
Chromium: 132.0.6834.210
Node.js: 20.19.0
V8: 13.2.152.41-electron.0
OS: Darwin arm64 24.5.0

Does this stop you from using Cursor

Yes - Cursor is unusable

[condor (Condor (Tee))]

Hi @gemyagouatc and welcome to Cursor Forum.

Thank you for the bug report.

Is that a custom trained/tuned model? Fine tuning might interfere with Agents functionality.

Could you post a Request ID with privacy disabled so we can look into the details? Cursor – Getting a Request ID

[gemyagouatc (Eugene Myasyshchev)]

It’s a standard gpt4.1, just goes proxied via litellm. ReuqestId is 7516460f-e140-44fa-b218-d548cd25eb7a, I’m not sure I can disable privacy, it’s a company plan.

[condor (Condor (Tee))]

Thanks for the update. As we can not look into a privacy enabled request it is hard to determine the cause.

I would like to help you resolve the issue, if you see a way to share more info it would help.

Additionally if you spot other issues here in the forum with litellm / gpt-4.1 over API key we may get from those perhaps more details to investigate.