I think that after cursor adopted the new billing rules, although it is said that there is no limit on the number of requests, I feel that the request rate has slowed down a lot. Is it because the model’s request response speed has been reduced
1 Like
No there has been no reduction in response speed as that comes from AI providers directly. I have had no slowdowns either personally
I tested it and found that there was definitely a decrease in intelligence at night. It was very obvious that the code was written haphazardly
Yes, tonight and today Cursors is unbelievable slow. It types a word per second! Quality was a bit worse starting yesterday eve
Which models are you using? Im running Claude 4 Sonnet and there is neither a performance/speed nor quality issue.
I used to have claude-4-sonnet always but I guess it unchecked somehow (might be me working in multiple workspaces). So I was working with “auto”. I am setting claude-4-sonnet back and hopefully “2x requests” doesn’t change anything since we are on unlimited now?
Regular claude 4 sonnet is 1 request and thinking is 2 requests if you think in old plan cost.
you get limits on claude 4 
in another post already confirmed that there is no relation between new price rule and speed
it totally different
why you get slow?
it model degradation quality
@danperks already confirm it
i also feel that the response has slowed down a lot
There might be restrictions during certain time periods. When I used it in the afternoon, I clearly felt that the response speed slowed down
that’s right
Yes, especially during some specific time periods, it’s very laggy
The model I use is the same as yours, but the response is extremely slow 
I see, the effect can have many causes, its not always easy to identify thats why I usually ask for a detailed bug report.
For example I have to turn off certain online backup apps on my machine as they slow down realtime streaming connections and it makes any AI crawl. With those off the AI is faster than I can read.
It happens also during busy times that Anthropic gets overloaded. However I get rarely affected by that as I chose my VPN exitpoint to a connection with great international routing.
Again, I’m using my personal Pro account all the time, but didn’t notice any slowdowns lately.
It’s not that the response is very slow in all time periods. It seems that only in some time periods does the response speed slow down, and it always replies word by word
Yes understandable, thats what I meant with parts of my comment. There are factors like Anthropics load, my internet provider traffic load (evenings can be really busy), my machines traffic load (backups etc.) and much more that can affect the speed.
Could you when the AI responses get slow again, check a few things and post in a fresh Bug Report with details?
- Open Cursor Settings > Network > Run Diagnostics (can run a few times, check Show Report)
- Check your devices Task Manager/Activity Monitor, anthing that is consuming a lot of CPU, heavy Network usage, heavy Disk usage. Try to close those apps temporarily.
- Check in Cursor the Process Explorer (under Help) to see if anything there consumes a lot of resources (cpu/mem/…). If its an extension you can disable it temporarily, If its a editor file, close it for a bit.
- Details from About Cursor
- If you can reproduce the issue, disable privacy temporarily and post the Request ID which the Cursor Team can have a look at.
- Does it happen with a new chat too, or on a long chat?
- New empty project also has the same slowdown?
Happy to go through the report and check things.
As Dan has stated clearly (in the link below) there is no effect of the plan or rate limit on speed of the model or quality of response. So it would be good to have a look at whats happening in details.
[Edit: added diagnostics]
This pricing business is far from idea. FAR from ideal. I’m on the Pro monthly plan. This is my experience so far:
- I was automatically added to the new plan. It was ok for a while but then just started hitting rate limits on every single model
- I switched back to the old pricing plan. I have Usage-Based set to $100/mo
- My new billing cycle started on the 24th (4 days ago). I still have 500 fast credits
- I’m using Gemini 2.5 pro like I normally do and it’s running on slow mode, like I’ve run out of fast credits.
- It shows me I’m running on slow mode and that I need to update my monthly Usage-Based to $200/mo
So, I pay $20/mo and have $100 Usage based credits and 500 fast credits but I’m running slow and have to update my usage based to $200 to use fast credits.
Not Happy.
1 Like
I have been using the new pricing model, and today I received the bill. In just 2 hours, I was charged $40, and I only used the claud4 Sonnet model. The so-called unlimited slow requests have not been triggered, and the explanation for this pricing model is still very vague. I have already turned off the new pricing model and disabled automatic subscription deduction for cursor, which is disgusting. It’s amazing how a software can do such disgusting things to users for profit
Yes, so I turned off the charge by volume，I hoped that the official will take a look at the feedback from community users and adjust the charging method or structure