After upgrading to v2.1.48 this morning, I experienced a total service interruption where all models failed to respond.
Symptoms:
Requests get stuck on “Planning next moves”.
Eventually fails with “Connection Error” or “Unable to reach the model provider”.
Critical Data Loss: After restarting Cursor to troubleshoot, the Agent chat sessions that failed to connect disappeared completely from the sidebar history.
Troubleshooting Attempts:
Checked local network (Connection is normal).
Restarted Cursor (Resulted in lost chat history).
Restarted PC.
Reinstalled Cursor.
Current Status:
gemini 3 pro: Seems to be working now.
gpt5.1-high (and others): Still returning “Unable to reach the model provider”.
Urgent Request: Please advise on a fix for the connection issue. Additionally, is chat history recovery possible? I have lost important work data in the vanished sessions.
Steps to Reproduce
Sending a request to any model
Expected Behavior
The agents starts working
Operating System
Windows 10/11
Current Cursor Version (Menu → About Cursor → Copy)
The entire GPT-5.1 series is unavailable, and the same applies to the full Claude series. They are all returning the error: ‘Unable to reach the model provider’.
I ran the command cursor --user-data-dir="%TEMP%/CursorTempProfile", and it seems to work normally in the new window. However, my original window is still unable to reach the model provider.
Below are the network diagnostic results:
If it still breaks with all extensions disabled, close Cursor and rename the profile folder: %APPDATA%\Cursor\User → User.backup, then launch Cursor and check.
You can work in a new window. To make the profile persistent, run: cursor --user-data-dir="C:\CursorCleanProfile» and keep using that profile.
I do not have the Nx Console extension installed, and the issue persists even after disabling all extensions. Fortunately, everything seems to be working correctly in the new window for now, so I will continue to monitor it.