I initially worked with local AI models but found them resource-intensive. They caused my system to overheat and slowed down performance significantly.
To resolve this, I switched to Cursor.ai’s premium plan, expecting smoother and faster operations. While premium requests are indeed faster, I noticed a surprising issue—some slower requests take as long as 3 to 5 minutes to complete, which is longer than what I experienced with local models.
This experience highlights the importance of balancing speed, resource usage, and system stability when choosing between local and cloud-based AI solutions. I’m using MacBook Pro M2 with 32GB
Yep, there is a trade-off between local models (uses your compute) vs cloud models (uses Cursor’s or their provider’s compute).
Cloud models will tend to be faster, unless you’re running a small model on a maxed out machine. Cloud models will also put a lot less strain on your system.
Regarding the slowness, is sounds like you ran out of premium requests. Cursor try and distribute slow requests as equitably as possible.
If you’d like to maximise speed, consider purchasing more premium requests or enabling usage-based pricing.
Using premium requests only speeds up the response time but does not resolve the system overheating issue. It feels like Cursor downloads and runs the models locally on the user’s system while giving the impression of using cloud models. This could explain the high resource usage and system heat generation — though I might be wrong.
Hey, in the processes, I see that the problem is related to one of the extensions, not Cursor. To confirm this, try booting in safe mode using the command in your terminal: cursor --disable-extensions
