Monthly limit on Pro

General
1

There seem to be a couple of mentions on the forums of people upgrading their Pro to 1000 rather than 500 requests per month. Is that a thing? Was that a thing? Will that be a thing in the future?

I’m on day 4 of using Cursor and I’ve created three-and-a-half fully functional projects and burnt through 250 requests already! Would gladly pay a bit more for a bit more.

I hope you’re not completely sick of being asked about this. I did try to do my forum research, and I even used the “Ask the AI” on the main site. It said:

To upgrade to 1000 fast requests per month, you can subscribe to Cursor Pro. Cursor Pro costs $20/month and gives you access to 500 requests with GPT-4, which is 3x faster than the default GPT-4 model over OpenAI’s API. This allows you to make more requests at a faster rate.

This seems somewhat contradictory; is it hallucinating? :slight_smile:

Thanks.

1 Like
2

Currently we only have the one Pro plan with a limit of 500 fast requests, with no way to upgrade.

We’re thinking about adding a Pro+ plan, in one of two ways: (1) usage-based pricing above the 500 included requests, or (2) fixed-price for 3000 requests.

Which one would you prefer? Our main concern with the usage-based one is that the incentives are not entirely correct (it’s better for us to have a monetary incentive to use the models more efficiently!), and our main concern with the fixed-price one is that it’s inflexible.

2 Likes
3

Thanks for your speedy reply! I guess it would depend on the price point (he said, unhelpfully :laughing: )

Personally, I’d almost certainly be interested in a larger fixed price subscription (with the slow pool fallback). Compared to PAYG on a personal API key, there’s something enormously liberating about not feeling you’re getting punished every time you want to ask another question! It might seem weird to “want” to end up paying a bit over the odds in quiet months but at least it’s a known fixed cost up front, no surprises.

Many thanks for the consideration and sharing the current thinking.

2 Likes
4

Is there an updated after almost a year? I am new user, and just for the past 7days trial, I’ve used 150+ requests, 1/3 of the whole, which indicates this is not enough for a month.

By the way, when the limitation is reached, would the personal API request be also limited to slow version?

1 Like
5

I am interested in the 3000 fixed price. It would be great to have the 200k or even the new 500k Claude option for composer. Thanks!

2 Likes
6

@dbsx ,

Here is some more recent information and research:

1 Like
7

also @arvid220u and @msc, after careful consideration, here are my thoughts(maybe naive):

I recognize that Cursor is a leader in AI-powered code editors, even surpassing Microsoft’s Copilot in capabilities. However, regarding model usage limitations, I believe the best approach would be to NEVER LET USERS REALIZE THE LIMITATION. Since even more general AI tools like Perplexity.ai don’t impose limitations on their pro models for basic usage.

The key is to let users focus on what truly matters when using Cursor - how they want to code with AI assistance, rather than worrying about reaching usage limits. By which I would rather there is an alternative pricing model for pro users(perhaps $25) per month, with which USERS NEVER NEED TO WORRY ABOUT LIMITATIONS. From a consumer psychology perspective, usage limits are likely the last thing users want to consider when utilizing AI during subscription.

Meanwhile, I think the philosophy of “making it simple or even transparent for users” isn’t just Microsoft’s approach with Windows, but also should be Cursor’s guiding principle. I believe it is, which I felt strongly when using Cursor. Then why not push it further, allowing users to focus more solely on their coding work?

I’m not sure whether canceling the limitation forever is or not a good idea from the perspective of a corporation’s sustainable development since I’ve never been a CEO or CFO. But you do, then what about just considering the recent popular case of an 8-year-old girl using Cursor to generate a simple “ChatBot” program:

"Do we want the girl to ever worry about usage limitations when she’s bought the service? Should she have to understand the concept of model limitations, what the token limitation acts for a model, or even what an AI model is technically? "

1 Like
8

Personally I prefer usage-based pricing. But that’s because I mostly use long-context, due to a larger codebase, so I run out of 10 long context requests quickly, but rarely run out of the regular fast requests, so I don’t need more of them. I think both options should be there. Also I would love to see more long-context improvements for Composer even if I had to pay more.

2 Likes