May I ask, where and when exactly was this “stated”? I cannot find it in official communication or docs.
BLOG: Updates to Ultra and Pro plans, June 16th, update June 30th
Existing users can choose to stay with the “500 request limit” method if they prefer.
FORUM: Cursor Ultra - A new, higher tier plan, built alongside the model providers
Existing users can choose to stay with the “500 request limit” method if they prefer.
BLOG: Clarifying our pricing, July 4th
No mention of old plan
If your US-based clients have any self-respect, they’d open up a class action lawsuit against Anysphere.
I think it’s that simple. Your company has subscribed to a batch of tokens at Claude AI company. Assuming 1000 users reach an agreement of 1 million tokens per day, your job is to allocate these tokens reasonably. If someone doesn’t use them today, you can assign them to other users. If they use a lot, you first need to set an average usage limit for each person. If the limit is exceeded, slowly restore the speed. Your official task is to integrate more excellent AI and optimize the rules of cursor development projects, so that we can call less AI and use these tokens more reasonably. Isn’t that the only thing? You are a resource profit margin company, much better than Claude Code Company. Their artificial intelligence connections are few and lack transparency, so they cannot choose! Cost cannot be controlled! This is their weakness!
I don’t think you can lose money in this way! There are only so many tokens in total every day. If he uses less, use more. If he used more tokens, then everyone should use them less until all the tokens are used up today. By controlling the water level in the pool, we can regulate the speed of token recycling
The most important thing is that the total number of tokens should be displayed in our account
([email protected]) is just a bot that keeps telling you to use the $0.04 per-request option in different ways. This bot is nothing but a time-wasting staller. When it realizes it can’t solve your problem, it leaves you with a final message: “forwarded to the engineering team,” and then disappears — no follow-up for months.
I am using the new plan that I renewed on 27th July, I was using it casually for some small tasks. But got this message recently “You have used 90% of your included usage. Consider enabling pay-as-you-go or switch to auto for unlimited requests to avoid interruptions.” What is this reminder as before I understand that I can use Claude 4 (thinking) as the burst and local limit refill after few hours. So what changed now aren’t the refills working anymore? The thing is it’s so messed up and really hard to understand what the new system is doing.
I was fortunate enough not to opt into the new pricing scheme, so I am still on the “grand father’d” plan but since I know it’s going away, I will unfortunately have to leave Cursor because I’ve also seen a sharp decline in quality and LLM laziness. This results in a lot more API calls, which would just be overall expensive. There’s something that changed in the background that has severely impacted Cursor’s ability to remember context, but also not get lost in endless Agent calls. I’m paying for all these changes that are wrong, and for that, I am not willing to subsidize a subpar project. I would much rather pay more and get quality, than have someone intercept my API calls and make the experience worse.
Where is the option to go back to 500 requests kept?
It’s not in any of the cursor.com / en / dashboard options
I wouldn’t even know about it without this thread and now you seem to be saying “oh, you had a window of opportunity to do that, sorry you didn’t hear about it on reddit”
It doesn’t exist, you just simply see the api calls and you’ll see “Included” beside them. They are completely obliterating the 500 premium requests notion even if you opted out of the new payment plan.
I’ve been using Cursor since mid last year and the latest pricing switch feels shady and concerning. They scrapped/phasing out the old $20 for 500 requests plan and replaced it with a vague rate limit system that delivers less output, poorer quality, and zero clarity on what you are actually allowed to do.
No timers, no usage breakdown, no heads up. Just silent nerfs and quiet upsells.
Under the old credit model you could plan your month: 500 requests, then usage based pricing if you went over. Fair enough.
Now it’s a black box. I’ll run a few prompts with Sonnet 4 or Gemini, sometimes just for small tests, and suddenly I’m locked out for hours with no explanation. 3, 4 or even 5 hours later it may clear, or it may not.
Quality has nosedived too. Cursor now spits out a brief burst of code, forgets half the brief, and skips tasks entirely. The throttling is obvious right after a lock out: fresh session, supposedly in the clear, I give it five simple tasks and it completes one, half does another, ignores the rest, then stops. I prompt again, it manages another task and a half, stops again. Two or three more prompts later the job is finally done. Why does it behave like a half deaf, selective hearing old dog when it’s under rate limit mode? I get that they may not want us burning through the allowance in one go, but why ship a feature that deliberately lowers quality? It feels like they’re trying to spread the butter thinner: less work per prompt, more prompts overall.
Switch to usage based pricing and it’s a different story. The model runs as long as needed, finishes every step, racks up credits and charges me accordingly. Happy to pay when it works, but why does the included service behave like it is hobbled? It feels deliberately rationed until you cough up extra.
And coughing up extra is pricey. There is now a $200 Ultra plan that promises 20× the limits, plus a hidden Pro+ tier with 3× limits for $60 that only appears if you dig through the billing page. No announcement, no documentation. Pay more to claw back what we already had.
It lines up with an earlier post of mine where I said Cursor was starting to feel like a casino: good odds up front, then the house tightens the rules once you are invested. That “vibe” is now hard to ignore.
I’m happy to support Cursor and the project going forward, but this push makes me hesitate to spend more and pushes me to actively look for an alternative. If they can quietly gut one plan, what stops them doing the same to Ultra or Pro Plus three or six months down the track? It feels like the classic subscription playbook: start cheap, crank prices later. Spotify, Netflix, YouTube all did it, but over five plus years, not inside a single year, that’s just bs.
Cursor used to be one of the best AI dev assistants around. Now it feels like a funnel designed to squeeze loyal users while telling them as little as possible. Trust is fading fast.
P.S. STOP SHADOWING PEOPLES COMPLAINTS IT IS SO OBVIOUS. I HAVE POSTED MANY TIMES AND ALL MY CRITISIM POSTS ONLY HAVE 1 VIEW AFTER DAYS OF THEM BEING POSTED???
I don’t use cursor or anything and I might not remember anthropic pricing clearly, but looking at your token usage, I count ~34$ API cost
So, we went from 500 → Unlimited → 225, bc y’all know the other models are borderline unusable.
The same
This. I lost all hope. Riding out the old plan as long as it exists, then quit.
I’ve long respected the work behind your product, but recent changes have made that difficult.
The Cursor user base isn’t just any audience, it’s made up of developers, architects, and senior engineers. People who think critically, understand the law, and know what’s acceptable and what isn’t. Many of us have been around long before the rise of AI and know how to adapt, evaluate alternatives, and make informed choices.
Earning trust from a community like this is no small feat, it means you’ve truly built something exceptional. But once that trust is broken, this same community can be unforgiving. Subscriptions will be canceled, word will spread, and the product risks being left behind.
Just something to seriously keep in mind.
And I have no idea what my limits are
Sonnet has gone from daily driver to “only in emergencies” and even then 5 requests were $1 apiece.
“at least $20” - what kind of agreement is that?
and in 7 days I’ve nearly used 50% of that $20 - mostly using one of the cheapest models available which used to be “unlimited”.
I don’t mind changes. I appreciate prices fluctuate. But I spend my time trying to guess how to optimize things.
It’s not my fault it decides to slurp up millions of tokens, my projects are not that big! If it didn’t rewrite the whole files from top to bottom perhaps that would help!
I made the mistake of getting a yearly subscription, the bait and switch was an unwelcome surprise.
I’ve also tried Augment Codes 2 week trial and found their results to be excellent.
Given that Cursors finance team have opted for sabotaging their product to their users detriment, should we begin advocating other products such as Augment Code to other users, at least until they return to providing the service they promised?
I’d be keen to hear what other users are trying, we need something else for when Cursor arbitrarily cuts us off with their pricing changes.
How to go back to the original pricing model?
You can’t. They remove it from the platform. People that were on it before it was removed were moved to the new plan without warning, communication, or agreement is what I’m reading in the threads.
