Extremely limited models now after update to Version: 2.4.21 (Universal), and slow/no response to prompt, showing “connection error: The connection stalled. Please try again”
Steps to Reproduce
I updated to latest ver., and, input new prompt, and this happens, so I check back to auto model list, found only composer 1.
Expected Behavior
before I update to 2.4.21, everything moves just fine, I get accurate feedback from prompt analysis, and have multiple choice of models.
I turn on auto, and switch to composer 1, same connection issue
For AI issues: add Request ID with privacy disabled
Request ID: 81a9465a-0848-4ce8-b138-d4a1965beea1
Connection stalled
LTe: Connection stalled
at $mf (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9096:4764)
at ova.run (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9096:8170)
at async qyt.runAgentLoop (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:34190:57047)
at async Wpc.streamFromAgentBackend (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:34239:7695)
at async Wpc.getAgentStreamResponse (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:34239:8436)
at async FTe.submitChatMaybeAbortCurrent (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9170:14575)
Sign out and sign in
Cursor Settings → Sign Out, then Sign In again. A few similar bugs after 2.4.21 were fixed this way: Persistent Connection Error
Check API keys and Base URL
Cursor Settings → Models:
Make sure “Override OpenAI Base URL” is turned off
Check if you have any third-party API keys (OpenAI, Anthropic, etc.). If you do, try removing them temporarily
Network Diagnostics
Cursor Settings → Network → Run Diagnostics. Send a screenshot of the results if the steps above don’t help
If it still doesn’t work
Try creating a new test folder with an empty project and open it in Cursor. If it works there, the issue is probably in the local cache for your project. As a workaround, you can try renaming the project folder (MyProject → MyProject_v2): Connection Error / Stalled Agent requests for hours
Let me know how it goes. I’m especially interested in the results from the first two steps.
I’ve changed several network, now it’s finally back to working after re-boot…
Left issue is AI model list not yet refreshed, weird thing is in my account page I can clearly see all those models in default settings, yet I cannot get those refreshed in my setting page..
btw, I re-do sign-out/in earlier right after 2.4.21 update due to an issue that I cannot even send the prompt, keep telling me to refresh the window. That, was fixed after sign out/in.
About the model list, it’s not a bug, it’s due to regional restrictions. OpenAI (GPT), Anthropic (Claude), and Google (Gemini) block access from some regions, so those models don’t show up in Settings.
The Dashboard shows all models available for the Pro plan overall, but locally in the IDE you only see the ones that actually work from your region, like DeepSeek, Grok, Composer, etc.
What you can do:
Use the available models. Auto will pick a good one automatically.
Try your own provider API key that works in your region: Cursor Settings > Models > Add custom model
Thanks for the advice.
I tried rebooting, tried changing HTTP/2-1. Moving the project folder to another directory helped—apparently the cache was building up and everything was crashing.
Adding a custom API is indeed a solution, but the experience of switching the cursor’s custom API is extremely poor. It would be great to optimize this to provide a better experience for pro users.
