Obviously this is an image of the new pricing page, where we are given the options for unlimited usage or 20x unlimited usage for those of us that warp the confines of reality and surpass the infinite…
All jokes aside, this all seems extraordinarily vague. I’ve even browsed through the docs and haven’t found much clarification yet at least. What does “unlimited” pricing mean in this context? Obviously there will be usage limits as it notes next to the “unlimited agent requests” text, but I can’t find anywhere what these usage limits are.
This leads naturally to a few questions on how to best use the new version. Do we hit the usage limits based on number of requests? Do we hit the limit based on tokens used? How often do these limits reset? Do the “1x request” or .5x or .75x request even mean anything anymore as it pertains to the new pricing model or are those just there for people that choose to revert to the old mode? Is there ANY built in MAX usage in the new “unlimited” counter or are those still purely charged per token?
Notable clarifications that could be helpful on things I have read the last few hours: I have seen random user comments in forums stating that upon reverting to the old pricing mode, they can’t change back to the new “unlimited” mode. Is this intentional or a bug? I wouldn’t want to switch back to accidentally get stuck there in case I prefer the use case of the new mode. I have also seen a user saying they tested max mode and it seems to have not charged them on the new unlimited tier for at least a few requests. (Both of these could be completely inaccurate but without specific explanation of any of these things in the docs or pricing page I’m not sure how anyone should know without everyone just testing any particular use case out for themselves).
I’d appreciate if future updates could include more information in the docs or pricing pages (wherever relevant) so I didn’t feel the need to make a new discussion topic for each new pricing shift.
But in the meantime if I could get answers on at least those aforementioned questions (hopefully from or confirmed by staff) about what “unlimited” means and how users might best utilize their usage it would be greatly appreciated.
If anyone else has already tested or confirmed answers on any of these topics or other user claims that feedback would also be appreciated.
That’s classic Cursor for you. Make everything vague and expect the users to come up with theories, then select the theory that puts users to biggest disadvantage. ;p
It seems I have exactly what I wanted. My 2x Pro plan from 7 months ago still active AND usage based pricing I added yesterday when my fast requests limit was hit and I no longer had access to Claude-4. I ADDED usage based pricing without losing my original Pro Plan. Still getting excellent value from Cursor by being able to add more requests to my legacy 2x Pro account and unlocking Max mode at the same time. I’m a happy camper.
We’re working on updating our docs with all the relevant details and answers to your questions.
To summarise so far, rate limiting should be pretty dynamic. The models you pick will continue to have a “weight” to them, which will effect how quickly you get close to your rate limit - more expensive models will get your there quicker.
The system should be resident to bursts of usage as well as usage over days, weeks and a full month. Like a health bar in a game, it will refill over time.
Based on our analysis so far, we think the majority of users shouldn’t see a rate limit often in standard usage!
when I checked usage what exactly does “included in pro” mean? am I not being charged extra for these requests now? I get 1500 fast requests per month with my current pro plan. (I stacked it early on) and when I went over I was charged additional usage. today all of a sudden I did not see my cost go up. My worry is im being charged elsewhere and can not see where
Why is it par for the course for cursor so go and make sweeping changes and then update documents later. It just seems so hap hazard.
To add insult to injury this just seems like a way to obfuscate usage. It’s the fabled fair usage internet policy all over again. We need transparency and real metrics of our usage pressure. At least with the limit we knew were we stood, now it will be anyone’s guess. If fast requests become dynamic with system load that’s great, but I suspect it won’t be. I suspect these pro accounts are now on course to become the free account performance which will force users to upgrade to the top package.
I’m not sure which of the 3 or 4 different threads about this I should put this in, but here, I choose this one arbitrarily. Definitely didn’t need to start a brand new one:
It appears there is a bit more clarity published in this. Unless I am mixed up or the docs are mixed up, it sounds like even on the new Pro plan, there is some version of additional usage based pricing you can choose to access if you want to?
why are we constantly below some Crash Bandicoot style boulder where you guys act like this stuff isn’t a business thing and we’re all bewildered in this day and age on the AI being confusing?
either you’re communicating in a disingenuous way the way the programs and cost expectations work, or you don’t know them. most of us don’t really care which, but while the improvements in some areas are exceedingly great, others are regressions that would take most of us 3 seconds to conceptualize in your tool. it becomes a challenging paradigm surely, but come on man. people don’t have issue with the “environment” and I really dislike you presenting it like API consumption is a confusing thing to track and estimate even in the scope of gen AI. API consumption and rate handling is your responsibility, and its not confusing in technicality, just your agreement with your model providers, which I can’t empathize with because you allow the narrative to just, fester. here we are again on the forums, screaming into the void because its a rolling release roulette with you guys. you refuse to set any other narrative, and it gets extremely frustrating when its clear hardware is not accessible for us to do the bare minimum in the areas people need care in - OUR WALLETS. SIMPLE THINGS WE ASK FOR LIKE HOW DOES MY MONEY LEAVE IT?
if you can track context enough to know the max the model allows, you can devise an average. from that average you can devise a rough cost, and advise in the excessive empty space in your application what this means to humans as humans are meant to ingest it, the same way you expect them to exhibit humanness in using your tool. You can spend the time to offer it, not the time to communicate what it costs me? The improvements were honeymoon, because the real problem was just shifted - here are other landmines even you as an established AI enthusiast, developer in industry, and relatively not stupid man you must hop over - because oopsies, it’s not our responsibility. Come onnnnn man. Have some empathy with the narrative point you’re choosing, and what it means for us.
Thanks for proving my point. How long did that take you?
At some point we get to hold Cursor to the same standard. Sorry my verbosity was too much for you in the scope of AI, that also confuses me that you come here for shorter or exhibit less.
I anticipated such a reply, & I edited while you were reading:
I am suggesting to you that there are ways to help you be a better communicator. You rambled that out & edited it repeatedly to ramble some more. Take some time to rewrite things & express the thought more clearly. I don’t say it fight, I say it to encourage better communication.
That is a fair point. However, in the time you took to make your point, you just robbed it of all utility. We live in a new age where some of this needs to be expectation. These systems we are auto coding on take hundreds of pages to describe, and in some of those are dollar signs we leak. While it is excessive, it also is not. If it feels that way to you, that should be an alarm. It’s not changing. It is that much and that annoying and that abrasive.
