i dont see a place to purchase pro+ only pro?
While I don’t mind the prices and love the premise of Cursor, the bigger the codebase gets the more errors you get. Be it not able to edit the code, giving new types of errors and apologies every 5 minutes or producing broken code on the simplest of edits.. When I’m hitting ‘Restore checkpoint’ the fair thing would be IMO to refund the costs associated to those requests. I’ve tried all the models with the combination of MAX mode enabled for them, there are some parts of my code where I can’t even ask it to put in a simple word onto the ui, without any logic behind it. Asking to optimize the code in the smallest steps produces even more bugs. At this point I’m giving up on that part of the code, no idea what I could do anymore.
1 Like
Yes, that’s the whole problem, he can go around in circles for 4 hours without correcting the mistakes. Then he will write off 20-30 dollars for nothing. I am ready to pay money if I clearly understand that I set the task professionally and he completed it on the first try.
There is definitely a change in how cursor is behaving now, it’s like working with a stupid version of it, like most people are saying. I’m now seriously looking at just downloading vs coder and run without cursor and that is a shame since I had high hopes for cursor and its future, not so much any longer. Everything is taking twice or three times as long and it just doesn’t seem to remember anything, its like it got Alzheimer’s.
You can pay $10 for Trae and get 600 Claude 4
2 Likes
I never use thinking mode because the of tokens cannot be controlled. I tend to avoid agentic mode so I can keep tabs on what is changed. I think its a mistake to give tasks that are too large to the AI. The AIs are not ready for this as yet. There is lots of false hope. The situation will improve within 1-2 years. Cursor should be more transparent with its policies. Token generation is not free, and most people don’t realize how fast usage increases. This is not a rant, rather a realistic appraisal.
I’m going to disagree with you, as that ability is already here. Claude Code allows you to plan out a change using a kind of iterative prompt engineering with AI, before pressing go and executing the plan. I find this approach does a better than doing individual small prompts. However, I have found that as the size of the project balloons, so do the errors and price, so taking a library approach is how I manage scope. Without it, the project gets too big and AI … it loses it’s mind, it seems. That’s been my experience anyway.
1 Like
Appealing to Cursor to Respect Annual Subscribers’ Rights, Rectify Unfair Refund Policy
As an annual subscriber of Cursor, I am deeply disappointed and outraged by the company’s recent actions. According to the user agreement, I pay an annual fee of $192, which translates to $16 per month. However, when I requested a refund, Cursor blatantly disregarded the agreement, treating me as a monthly subscriber and calculating the fees for used months at $20 per month, which is clearly inconsistent with the annual fee I actually paid. This is a severe infringement on the rights of annual subscribers.
I hereby formally appeal to Cursor to uphold the spirit of the contract and calculate refunds fairly and reasonably based on the annual fee. As a reputable company, you should respect and protect user rights, rather than adopting this lack of transparency and disregard for users. I urge all annual subscribers to join me in using rational and factual methods to defend our rightful entitlements, compelling Cursor to re-establish a fair and reasonable refund policy, and restore user trust.
Our strength comes from upholding principles and persisting with facts. Only then will this company be held accountable for its breach of trust. I will take responsibility for the content I have expressed here, and as a reputable company, Cursor should also have the courage to be accountable and respect the legitimate rights of annual subscribers!reasonable refund policy, and restore user trust.
Our strength comes from upholding principles and persisting with facts. Only then will this company be held accountable for its breach of trust. I will take responsibility for the content I have expressed here, and as a reputable company, Cursor should also have the courage to be accountable and respect the legitimate rights of annual subscribers! We cannot allow such blatant violations of user rights to be condoned. Let us unite to protect the rights of annual subscribers!
I followed your advice and got it today. it’s so good, I won’t even miss Cursor, especially for vibe coders like me. Thank you, you’re my lifesaving angel!
Anthropic hired back Boris Cherny and Cat Wu — just two weeks after they left to work at Anysphere, the developer of Cursor.
Cherny and Wu were two leaders of Anthropic’s coding product, Claude Code
3 Likes
I’ve been using Cursor now for three months and love it!!! I’ve been a huge advocate and felt like it was easily the best in the business; however, as a smaller creator and student, the cost analysis of these platforms matters a lot. Before, on the student plan, I was rarely hitting limits and having no issue until now… I had a little bit of a lull where I didn’t use the platform and came back 2 days ago. In 2 days with moderate daily usage (2-2.5 hours), I managed to hit the limit on the pro plan and thought no big deal I’ll just pay for extra usage like I’ve always done (and never surpassed $25); however, today, within the time span of 4 hours, I have managed to use almost the entire pro plan in what felt like minutes, with prompts costing up to $2 per prompt, which is absurd. I’m not sure if this is a bug or just the new pricing platform, and I hate to switch, but if nothing changes or it’s not on a 500-prompt usage basis (with lower prompt usage for Claude), then I’m not sure if I’m going to stay even with a free account.
Cancel my subscription — I’m switching to Kiro.
This new pricing structure will be the death of Cursor.
6 Likes
Obviously, in the initial subscription of 500 requests, 500 fast requests imply that the user pays for the purpose of calling the most efficient and expensive model. As a side effect, autotab and small models (such as cursor-small) can be used in large quantities. When I subscribed, there was not even effective agent mode tools and jupyter notebook support (although it is still inefficient at present). It must be admitted that some improvements permanently provide the convenience of cursor as an IDE. I can understand and am willing to accept a certain price increase for this (because the effective agent mode obviously requires more context and tokens). But when it switched to the so-called unlimited requests (actually unlimited requests of the free model), it did not immediately declare and announce that they were actually switching to a traffic billing model (the original max part) and how to charge, and did not even notify me (until a message that I had saved $129 stuck my work). Even the public, which is now worth at least $20, still did not tell us how much it can be used. In fact, no one wants to work without knowing when they will be stuck. The subtext is that you should all use my max mode, and max should become the mainstream (from here on, the contradiction between subscription fees and subscription content arises, We initially chose a planned limited model instead of an unplanned billing model, which aroused the disgust of most people even though it was cheaper). Another hidden cost is that LLM’s answers or modifications are always imperfect and need to be reviewed (because most people are engaged in cutting-edge and difficult work, but cursor seems to position their user base as students who want to complete programming homework but are unwilling to actively learn knowledge). In the past, based on the fixed request calling method, a request can contain more context and make more modifications, which is better and more cost-effective for users, but after switching to the billing model, its weaknesses are exposed. We are unwilling to pay for these trial and error and unnecessary AI’s recklessness unless they provide a more efficient model that only completes what I stated. It is obviously unwise to keep subscription before the situation improves, but it is obvious that most people are actively looking for alternative products (because their work is stuck and they have to do something else). It is an honor to witness the killing of a great idea. I used translation tool 
.
I agree, it’s crazy how we don’t know when we will run out, so there is no way to plan or pace our work…
Hey Dan
i saw my dashboard usage summary, the data is
| Model | Input | Output | Cache Write | Cache Read | Total Tokens | API Cost | Cost to You |
|---|---|---|---|---|---|---|---|
| claude-4-sonnet-thinking | 114,793 | 1,113,835 | 14,573,618 | 161,973,629 | 177,775,875 | $120.30 | $0 |
| Total | 114,793 | 1,113,835 | 14,573,618 | 161,973,629 | 177,775,875 | $120.30 | $0 |
I am wondering why the cursor is so expensive? if i use the claude api directly, the token fee is almost cost $20, why cursor is 6 times than claude?
I don’t use Cursor much, but I’ve already reached the limits when the subscription is still 10 days away. The new pricing system is very bad… At least leave slow requests… Auto mod is very bad, I want to use sonnet 4
2 Likes
There is a link somewhere in the forum, the only problem is that you have to pay the full price, you can’t just simply upgrade.
If you have pro annually like me it will be painful to upgrade to monthly pro+, also because they don’t offer an annual version of pro+ as I understood.
1 Like
I’m just an insignificant pro annual subscription user. If cursor forces all its users to adopt their new pricing plan, I will definitely cancel my subscription because I can’t accept an opaque usage billing model. 500 times a month, simple and clear, isn’t that great? It has also attracted a large number of users for cursor. However, the new pricing plan I feel that cursor has already lost a lot of users. It’s just that the fermentation time is still relatively short now, so it’s not obvious. The real results will come out in two or three months
Since the price change, the usage has inexplicably exceeded the limit, and we have received bills inexplicably. The charges are acceptable to us consumers, but are they providing us with reasonable and appropriate results? The so-called infinite generation now cannot be used due to continuous interruptions, and the generated content is unavailable. I have sent you feedback via email but have not received a response. I now want to restore the previous 500 high-speed requests. I hope the official can solve this issue. There is no button in the advanced settings of my account to restore the previous mode.