Pricing Megathread and Q&A

Hey all. We are consolidating discussion around our June pricing update for Pro into this megathread. We want there to be a place for you all to share feedback, as well as a place for our team to help clarify and answer your questions.

In case you missed it, we wrote a post about the pricing change and answered common questions about how the pricing works. One the biggest pieces of feedback from this subreddit was to add better usage visibility in the editor and dashboard, which we have since shipped.

New pricing posts will be asked to move in here instead so we can better respond and answer questions.

We’re going to continue listening to your feedback and finding ways to improve the product and pricing experience in Cursor. We will update this post with frequently asked questions as they come up.

Q: Was the previous Pro pricing removed for existing customers?

No, existing customers still had the ability to opt-out and continue with request-based pricing. If you are still on that plan, nothing has changed. We will eventually sunset that plan in favor of our current pricing, but we want to make sure we honor the existing billing period (e.g. for the remainder of your yearly purchase). We’ll be emailing customers directly, so you will hear from us first.

Q: I wasn’t able to opt-out, can I go back to the old pricing?

New users are not able to go back to our old pricing, but for existing users wanting to use request-based pricing until we sunset the plan, you can reach out to our team ([email protected]) and we’ll help smooth this transition.

Q: I purchased a yearly plan, am I able to stay on that pricing for the rest of the year?

Yes. The pricing you purchased will continue to be reflected for the duration of your billing period.

10 Likes

Thank you. Could you clarify the rate limits in more depth? As far as I understand, if I run out of my plan’s allowance, I either go with usage based pricing or the rate limits. Are the rate limits identical across plans? Or does ultra have higher allowance (after exhausting the included api costs)?

I would appreciate more unified model information, e.g. when new model is added to cursor to be used, to be automatically listed in the docs page, currently o3-pro and grok-4 are missing in the docs! Also some models within cursor are missing context and price information (when hovered in chat selection)

Let me link a few threads I would appreciate feedback on (related to latest changes):

Q: I purchased a yearly plan, am I able to stay on that pricing for the rest of the year?

This is lie. I purchased the yearly plan, you switched me without even my acknowledgment to the new plan. Yes, you gave me all the money back without even offering “I’m sorry” when I switched to 60 plan to test it. But this is total lie. And also not in line what you wrote above that you can go back now to the old plan (you can’t, check the forum, zillion threads about that) and saying at the same time you will cancel that option one way or another.

8 Likes

Thank you for the follow up Dan, as one of the folks who was upset I would like to say that I have already uncancelled now that you guys have refunded me the unexpected charges. Now that I know what it costs and can plan for it I am okay with how you guys are going about it. I appreciate what you guys do for the community, I think all we could ask for at this point is just very clear, honest and open communication. Banning everyone in the Cursor subreddit who were calling out these practices did far more damage to everyone’s trust in Cursor than if you guys had just ignored the posts and just kept cooking a solution in the background. You may wanna have a talk with your subreddit moderators and maybe teach them that silencing and banning people who may have been upset and posted is not the way to go about the solution at all and probably caused you guys way more headache than necessary.

For example, I now know not to even bother looking at the Cursor subreddit due to the repeated instances of people just getting not only their posts removed but banned entirely.

6 Likes

The fundamental issue I have with the new plan is that on Auto, my Cursor experience becomes somewhat worthless. My typical process is Claude-4 for the initial request, then 3.7 for test reconciliation and debugging. Now that I’ve run out of credits for the month, I’m forced into Auto Mode, or what I like to call “Stupid Mode.”

The quality of the result significantly declines. I don’t know what it’s, but Auto seems to get confused and ends up asking me a lot of questions, so many that it seems to forget what it’s doing. For example, it produces a To-Do list, starts on the first step, but then never makes it to the second step. After a while, it feels like it’s in a loop, where fixing one test breaks two others, and fixing those breaks the original one. Back and forth until the inevitable GIT discard and try again. I’ve stopped developing with Auto Mode, unless it’s something relatively straightforward.

What I would like is a $20 usage-based plan, but with the ability to add credits in $10 increments as needed, that roll over. (EDIT: Or a $10 plan that auto-rebills when I run out of credits to some limit I set, like US$40/month. Using tiers doesn’t make sense where a few sizes fit all use cases. A fundamentally transaction-based platform should have a subscription model that is also transaction-based, too, shouldn’t it?) I’m personally distressed that usage-based is being phased out and replaced by the crippled Auto Mode.

13 Likes

I just want to focus on coding—I have no interest in getting involved in your internal power struggles. Back when pay-as-you-go pricing was available, that was the happiest and most seamless experience I had. Later, you added a so-called “toggle button,” which turned out to be completely fake—the experience didn’t match what was promised at all.

Now, you’ve removed it again. What exactly are you trying to achieve? If you can’t genuinely improve the user experience, please don’t make grand claims like “unlimited usage.” In some less-developed countries, $20 is not a trivial amount. If you believe your service is superior to the competition, then confidently raise the price—don’t keep toying with users.

Over the past month, I’ve noticed occasional slower response times and even queuing for simple requests, despite having used less than 112 requests. Now, with the new pricing model, you claim I’ve used $30 worth of API usage. Honestly, this feels disrespectful to users and is slowly draining the enthusiasm I once had for products like Cursor.

17 Likes

Here’s my feedback:

  • I used to pay $20 plus about $120 a month and could do everything I needed to do. It was perfect
  • Now, it’s a s**tshow, my cost has at least doubled, and I’m actively looking for alternatives
  • If I switch to the agent + auto I get assigned random LLMs. Sometimes they are ok. Sometimes they complete mess things up and I have to undo everything but you don’t know that until after the fact. I don’t actually know which LLM I’m using so the results differ all the time. It’s pretty much useless
  • Because the agent + auto is so bad, I’m basically forced to used a different solution. I choose sonnet + agent as it’s the best and most reliable which has, at least doubled my cost
  • Everything was working perfectly before. You could have just done a price increase from $20 - $25/mo and I’m pretty sure 99% of people would have been ok with it
  • This whole s**tshow sounds like the finance team making product and tech decisions. Fire whoever that person is because they are a cancer that will kill your business
  • You can literally install VS code and use any one of a dozen plugins, add your own API keys, plug into openrouter and use free LLMs and off you go. There are pros and cons to this but I’m pretty sure the pros far outweigh the cons of this pricing situation.
16 Likes

I also opted out of subscribing to ProPlus, I can accept the price increase, but I can’t accept opaque consumption, what I need is a stable and usable sonnet, not a plan to reset after a week of use, and I can no longer use it.

6 Likes

My main issue here is the costs that are being associated to certain models just don’t stack up with what those models advertised costs are (even with a margin) - in particular OPUS, $15/m input, $75/m output, this is the figures for my opus usage - unless cache tokens are having a big impact here:

Model Input Output Cache Write Cache Read Total Tokens API Cost Cost to You
claude-4-opus-thinking 408,614 132,355 5,704,046 33,004,329 39,249,344 $205.38 $0

Clarity around this would be great, no point offering $xxx dollars of value if the pricing isn’t clear.

4 Likes

they +20% margin when calc api cost

4 Likes

you put it well. it’s just another flavor of internal something I’m paying for when I shouldn’t have to. most companies don’t do this type of stuff on forums. they have MOTDs and for whatever reason hold themselves to a higher standard. not sure why we continue to get shoveled new models and weird agentic implementations at the rate I am, because that’s not the issue with the product currently nor was it 6 months ago. it literally feels like Path of Exile trying to unpack anything - we’re buffered from almost all useful consumer protective information

1 Like

My original post 13 days ago was shadow banned so I am posting here.

I got absolutely DESTROYED by this new pricing model. I did 3.4 million lines of code last month for less than $150 using Sonnet thinking 3.7 for most of it and then 4 at the tail end AND I used MAX mode last month for about 25% of my work. I was about 11 days into a new billing month and I already hit $150 in costs at less than 30% of the same usage. Half requests are getting charged MAX pricing practically. I had no idea I was getting charged this high of pricing. Same project. I’m absolutely LIVID! I never agreed to this new pricing structure. Is this even legal to have secretly switched to a new pricing plan without proper notification? No email, no warning in the app, nothing. I had to find out on the forum after seeing my costs skyrocket out of nowhere! Am I missing something here?!

15 Likes

can i got refund for an ultra sub? refill rate so low now that i can’t do meaningful work with it, barely touch it after 2 days sub

1 Like

Since you keep moving the goalposts, making up different pricing schemes and changing conditions every few weeks, I screenshotted this commitment, so that hopefully your own lawyers will keep you accountable. If not, EU laws will certainly hold you accountable and the EU might come after you for repeatedly changing contract conditions.

Also, there are contradictory statements about whether slow requests are still available under the old pricing, with some users reporting that unlimited slow requests are not available any more or somehow more restrictive than before.

As to your communications performance, it is absolutely abysmal. First you make changes without proper communication and then you hide posts that call you out on that. Shadow-banning posts on your forum and blaming it on an “automated system” is a really pitiful excuse. Multiple neutrally worded opt-out posts were shadow banned, including my post which you reinstated, but locked down. You programmed that “automated system”; fix it!

Developer trust in your company has been thoroughly undermined by your own actions and the VCs that might have pressured you to make those terrible pricing plan changes will not be happy when they see developers cancelling accounts in droves.

10 Likes

I came back to a generation that, after explicit instructions, disobeyed me and ran python terminal commands. I understand there is a hard stop, but it runs others just fine and this is clearly an environment symlink mess up.

Look, I’ve wasted TONS of generation on this very thing. I have no recourse for it. I’m actually fed up. I’ve unsubscribed and refuse to support this company any longer. At all ends of the candle I said it was inappropriate, lacking, and not protecting of consumers, gasp, here it is. Come on guys. Seriously. I’m currently trying to figure out what insane alias it did in the 5 seconds I looked away so that a new Cursor doesn’t open EVERYTIME ANYTHING EXECUTES. This python environment thing is a huge problem among so much else, and this all just drives API use. I’m not giving them any more money. I’ve had it.

EDIT: an 8B local model, when tested, and smaller agentic models than that, do not deliberately do something i tell it not to. they all respect boundary conditions quite well. it’s time to start asking why it continues to be coaxed to do certain things completely outside the bounds of expected embedded context given user query Q. It’s just devolved from a cool system with promise to a complete lootbox shell of whatever it is. This is not how AI stewardship looks.

2 Likes

Did you try to opt out of the new pricing model (if you even heard about that in time)? You might still be able to go back to the old pricing by emailing them directly. - Did you request a refund? They promise some refunds for excessive charges during that time period.

or we hook an employee’s wallet into a normal average user’s account and let them learn Cursor in a natural bumpy way they’ve designed, or allowed (not one to care much about intent, seems to scapegoat most bad behavior):

and then see how quickly things change. it’s not fair and it isn’t right. I don’t know how this requirement of me doing mental gymnastics to assess value is considered normal. 80 out of 100 people stumble into a high bill or lower expectations and a higher bill.

1 Like

I wrote to the mail, as you wrote yourself. Please transfer to Opt Out. There has been no answer yet. Waiting.

2 Likes

Yep. On top of the, at least, doubling, of my monthly costs, I too, got smashed by the new pricing. I think I burnt through my 500 credits in a couple of days. I was trying to work it out, testing old versus new, this config versus that config. I think I blew threw $124 in a day. I also got charge ~$80 something out of nowhere. No idea what that’s for. I’ve email Curor asking for a refund.

2 Likes