same for me, if i need to pay 200$ i will go private, deploy a server with a GPU with i nice opensource for code model and all the data is not shared in the cursor api’s.
A friend of mine just threw 12k to do exactly that, another 9k. I’m waiting for them to test it. I think Cursor doesn’t understand that the problem for a lot of people is not in money, but in fact how Cursor wants that money…
Hi! I’m interested in Cursor Ultra plan, but I’m working in a large codebase. I know Pro plan limits codebase indexing to 100k files, could I index a 500k+ file codebase with the Ultra plan? This would make upgrading definitely worthwhile!
If not normally allowed, would love to work with someone to figure something out that works. I want to subscribe!
Due to the length constraint, let me continue implementing the remaining components in the next response. This is a comprehensive implementation and I’m making excellent progress. Let me continue with the backend implementation by focusing on the critical remaining pieces to make this fully functional MVP.
I never got such response in Claude Code. Using Sonnet 4, with quite a small task. Company account meaning - worst case - it would stop after 25 tries. If this is not throttling I don’t know what is.
To make it as absurd as it goes: I am getting since a couple of days notifications that 22nd of July is the day when I am projected to go without limits. I didn’t use Cursor at all since then with my own account. Turned it on now on my laptop, wrote a single message, and without it doing anything got the message that I am out of my credit:
This is as absurd as it goes, and it is not about usage anymore, this is either as buggy as it goes, or you guys are really playing dark patterns here. Cancelling my subscription, Claude Code does magic, and gonna try Codex as well.
holy moly, do you code a new operating system? what do you work on that is that huge?
changxiaoqiang,
I’m currently in a similar situation and was wondering if you could kindly share how you managed to cancel or request a refund.
Did you contact support directly, or did you take any additional steps to get the unused portion refunded?
I’d really appreciate any advice or steps you can share. Thank you in advance!
I canceled the subscription service and sent an email to their support team requesting a refund for the unused portion of this year’s payment. I’m currently waiting for their response regarding the refund…
The opaque consumption is the killer
I’m a new Cursor Pro user and I’ve been using it for 2 months now. What confused me was how the unannounced pricing policy change in the app, or the way Cursor communicated it to me. As a user, I don’t check forums and subreddits to see updates and news. So It would’ve been better if I’ve been shown a modal that clearly says “We raised our api pricing, keep it in mind” and only remove it when I click “I agree”.
Last month, when I can credits in the dashboard I barely used 180/500. Keep in mind that I used it for various projects intensively. So I assumed that even if my credits got cut in half I could still be within the limit. But now, used mainly sonnet 4 for 3 weeks and o3 for 4 days and somehow I ran out of credits.
Users are entitled to be notified about costly models and the way Cursor did it was unprofessional and greedy (“oops, you ran out of credits (which can’t be shown like before), pay more or switch to PAYG pricing”)
If Cursor was a B2B IDE from the start, the pricing increase would be understandable. But weeding out the paying individuals is seen as nothing but greedy.
I started using cursor T3 2024. Early days were fantastic, 20$/month for unlimited intensive usage felt too good to be true.
Then, fixed 0.04$ per request, 20$ per 500 request felt reasonnable and fair. Easy to understand, easy to follow and monitor.
Today, I’m able to burn 20$ in less than 6 hours. No control on what’s happening. I’m not even adding any context. Model runs on its own, based on my request and can go up to 2$ for a single prompt.
I’m working on a little chrome extension which have approx 5k lines of code. wtf ?
This is ridiculous. I’m out.
Edit: I should actually thank you. This non-sence gave me a good reason to look for a sane alternative. I’m pretty happy I have found Trae which reminds me of fair early days.
It bothers me when the forum mods reply to posts like yours and say “It all makes sense! You fit too much stuff in the chat context!”
It’d belittling. It’s like they’ve lost touch with people who write code for a living.
With these new pricing changes, nearly everyone has the same story of “I hit the monthly quota in a few hours of the first day”
That’s NOT normal. I paid $20/m, and was forced to upgrade to $60/m just to not pause mid-day.
And saying stuff like “just use auto”. I sometimes wonder if the employees of startups are completely out of touch, they don’t have side projects, they don’t build real things and don’t understand that ‘auto’ mode is so weak. I would rather copy paste my code into grok or chatgpt and ask for it to change stuff than use auto. The only reason I use cursor is it’s slightly more convenient than pasting code into a chat interface.
yea not gonna lie this new pricing is ■■■■■■■■
you’ve lost my money
Cursor being an expensive product is okay, but if 2 days of moderate use is going to use 70 dollars of my 400 dollar usage, then the “ultra” plan is not very ultra and should either be named as such or should be priced to actually be… ultra lol
waiting till the 500 fast request cames back…
I want to switch back to the old mode, can it be achieved, I do not have this option now
I’m sorry to say this but cursor is atm in a bad position. First aws anounced his first ide named kiro, and seems google added agent mode for gemini Gemini Code Assist’s June 2025 updates: Agent Mode arrives
Confused by additional charge what does “compute usage” mean?
Hey everyone,
I recently got charged beyond the regular $20 Pro subscription and only understood why after reaching out to support.
They said it was due to “usage-based charges for compute resources beyond the included quota.”
Here’s their response (screenshot attached).
I honestly didn’t understand this at the time. There were no clear warnings or live indicators during usage – just surprise charges afterwards.
I thought I was on the $20/month plan, and nowhere was it clearly shown I was exceeding anything.
A few questions:
-
What exactly counts as “compute usage”? Is it just token count? Or model-specific?
-
Why isn’t there a visible, live tracker or warning when you go beyond the quota?
-
If I bring my own API keys for models like GPT‑4, Claude, etc. – will this drastically reduce costs? Or will Cursor still charge me for compute?
Would appreciate any clarification from the team or community. I really want to keep using Cursor but need to better understand how to stay in control of spending.
Thanks!
I’ve noticed these patterns too, as well as the teeny-tiny (2x usage) for Claude that was what consumed most of my credits this month. Seems to me that since Cursor is built on VSCode, and uses third-party LLM providers, that the real product here is user experience. And the real value for Cursor (the company) is a happy user-base. It’s a little early for Enshittification - Wikipedia
In documentation there used to be a table with models and its prices or request usage per model.
Where can I find that?
I just need to know how each model consumes request usage so I can plan which model to use to optimize token/request usage.