Whenever I try to use models like Sonnet 4.5, Gemini 3.1 Pro/Flash, Grok, all is working extremely slowly showing “Taking longer than expected…” and then gives output and then when it gets to Planning again, then again “Taking longer than expected…” for 2-3 minutes for each Planning, working extremely slowly, completely unusable.
Steps to Reproduce
Select any AI Model as mentioned and send a request, takes so much time to give output.
Expected Behavior
Should work instantly without showing “Taking longer than expected…” specially for almost all ai models.
Hey, this is a known issue. Several users are reporting the same thing. I see you already tried disabling HTTP/2 and ran network diagnostics. Let’s try a few other things.
Check the size of your state.vscdb file. Open PowerShell and run:
Then install sqlite3 with winget install SQLite.SQLite and run:
sqlite3 "$env:APPDATA\Cursor\User\globalStorage\state.vscdb" "PRAGMA journal_mode=DELETE; BEGIN IMMEDIATE; DELETE FROM cursorDiskKV WHERE key LIKE 'agentKv:%'; DELETE FROM cursorDiskKV WHERE key LIKE 'bubbleId:%'; DELETE FROM cursorDiskKV WHERE key LIKE 'checkpointId:%'; COMMIT; VACUUM;"
This will clear your chat history in the UI, but agent transcripts stay in ~/.cursor/projects/.
Test in an empty folder. Go to File > Open Folder, pick an empty directory, open a new chat, and type something simple. If that’s fast, the issue is tied to your project.
Start fresh chats more often. Long conversations grow the context, and each response can get slower over time. If you notice it slowing down, start a new chat.
The team is aware of this pattern. I’ve passed your Request ID along. Let me know how the state.vscdb cleanup goes. That’s been the biggest fix for other users hitting this.