Fix: Cursor AI Responses Are Extremely Slow
Quick Answer
Switch to a faster model: Settings > Models > select GPT-4o-mini for tab completions and fast chat. Keep Claude for Composer tasks where quality matters more than speed.
Quick Fix Summary
| Issue | Slow AI responses |
| Fastest fix | Use faster models for completions |
| Use this page if | Chat responses take 10+ seconds to start |
Symptoms
- !Chat responses take 10+ seconds to start
- !Composer hangs for 30+ seconds before showing results
- !Tab completions appear after you've already typed past them
- !Progress spinner runs indefinitely
Step-by-Step Fix
Use faster models for completions
Go to Cursor Settings > Models. Set tab completion to GPT-4o-mini (fastest). Use Claude Sonnet for chat. Reserve Claude Opus for Composer only.
Reduce context size
Close unnecessary tabs. Use @file for specific files instead of @codebase. Smaller context = faster responses.
Check your internet connection
Cursor sends requests to cloud APIs. Run a speed test. If latency is high, try a wired connection or different network.
Restart Cursor
Sometimes the language server gets stuck. Cmd+Shift+P > 'Developer: Reload Window' or fully restart Cursor.
Frequently Asked Questions
GPT-4o-mini is the fastest for completions. GPT-4o is a good balance of speed and quality. Claude is slowest but highest quality for complex tasks.
Yes. Larger projects take longer to index and provide context. Use .cursorignore to exclude node_modules, .git, and build directories.
Related
Weekly Signals
Get the next fix, switch, or warning before it hits your build.
Join builders getting the community signals, fix patterns, and tool shifts that matter before they show up everywhere else.
Follow the signals →