I have used Cursor on some internal projects for days. And I never noticed that I have a choice on using what kind of models.
And I have sent like hundreds of requests, including easy tasks like creating a simple dialog, and complex tasks like create a complicated code for a complicated UI. But I was never asked things like “Do you want to go fast requests”.
Can somebody explain this to me? Do I have control on my own? What if the quota became 500/500 and I don’t pay extra money > What would happen?
In Cursor, we have two groups of models: “premium” and “non-premium”.
Premium models are the smartest, but most expensive to run. In Cursor, you get 500 requests to premium models that are “fast”, and are executed instantly. Once you run out of fast requests, you then can use “slow” requests, which may be put in a queue to be completed.
If you want more fast requests, you can enable our usage pricing toggle in the dashboard, and pay $0.04 for every premium request outside of your allowance, knowing they are fast and will be executed immediately.
Our non-premium models are unlimited and always fast, and while they are not as smart as our premium models, can often be used for a lot of what you may be working on in Cursor!
Your requests are always fast to start, until you run out of your usage allowances, at which point they become slow.
The premium models are the ones that perform the best on benchmarks, so usually are the best at figuring out what you want them to do, and writing the code with the fewest bugs or issues in as possible.
However, they are definitely not immune to these issue, and can still have these issues if they are lacking the right context they need, or are trying to implement something too complex in one go - in this case, breaking tasks down into smaller ones, and doing them one by one, often works much better.
I have been heavily using Cursor to create some complicated projects (involved with many files, npm packages, …), and though I don’t remember the exact number, I am sure I’ve sent more than 49 requests to the agent. So how do you measure the count “49”? Just wondering.
Your usage refreshes every month, so it may be your usage has reset already.
However, each request to the agent composer uses 2 fast requests worth of usage, whereas the normal composer chat and CMD+K uses 1 fast request.
It may also be that the dashboard is not showing your live usage, and may “catch up” in a few hours to your real usage, if you have used a lot of requests very recently.
