Hi, I just read the documentation they updated today, and I don’t like what I saw.
So they raise prices starting September 15, and on top of that, they charge you for the “Auto” option even on the paid plan, if I understood correctly. They don’t even clarify which models are used in Auto.
I even sent an email asking specifically which models they use in the “Auto” feature, but they skipped answering.
I think they should clarify which models are used in “Auto” mode.
If at least they charged you a million dollars, and that was their policy, that would be fine — no one is obligated to pay — but at least users would know what they are consuming and paying for.
But now it seems they are making everything more expensive without telling anyone, and charging you for everything.
This is from their documentation today:
“Starting at your next billing renewal after September 15, Auto consumes included usage at the following API rate:
Input: $1.25 per 1M tokens
Output: $6.00 per 1M tokens
Cached Input: $0.25 per 1M tokens
Both the editor and dashboard will show how Auto consumes your purchased usage. If you prefer to select a model directly, usage is incurred at that model’s list API price.”
So they are even going to charge you for the input in “Auto” mode on a paid plan, even if you don’t know which models they are using — they could even be using a free agent.
My opinion after this — even if it doesn’t matter — is that other agents/editors are getting stronger, and if this keeps going, Cursor will lose a lot of users. That’s very sad, because I think Cursor is the best code editor today, but at this pace I personally doubt it.
Sorry if I sound whiny, but I can’t believe what is happening.
I really hope they lower the prices or at least make things clearer and not “under the hood,” taking advantage of all the people that trust Cursor. I hope they change something in the next days.
If not, this will be very sad and disappointing, for them and for us, I think.
Starting at your next billing renewal after September 15, Auto consumes included usage at the following API rate:
Does this mean if you get an annual membership before September 15 you get unlimited Auto for the length of your subscription before renewal? (I know that’s what it says, but is that actually what Cursor means… who knows)
I have the same doubt as you. Maybe the “Auto” feature will be unlimited if your subscription ends before September 15, or maybe not. I’m not sure which of the two it is.
hmm that is bad news to me, if auto model mode require budget also instead of others
then I felt sad
IF that will be truth about that, I need to purchase extra request via Usage-Based Pricing
but, this isn’t good to me
well cursor provide before api pricing based charging introduced
like gemini flash, cheap model provided to subscriber without charging
I hope if cursor okay, at least low cost model usage provide to us
At this point, is there a reason anyone should consider Cursor? Like, there are so many cheaper alternatives that do the same or even better. Claude Code, OpenAI Codex, Windsurf, etc., etc.
People keep saying this. And yet here they are, on Cursor forum, still using Cursor. Why haven’t you all left and why haven’t Cursor died as a result? Really makes one wonder if those alternatives are actually cheaper and better
Idk honestly why I’m here, just for the vibes I guess. Already refunded my subscription and use Claude code instead for a few days now. I don’t have cursor even installed anymore.
Maybe if you can pay $200 a month, Cursor is still worth it—though even then I’m not sure it really feels truly unlimited. On the $20 plans it’s like a casino: every day you’re unsure whether you’re spending your money wisely or just getting unlucky. I’m not blaming the Cursor team; what came before was basically investor-funded promo, which still feels weird. But yeah, there are alternatives now.
Anyone know if this pricing listed is what is currently being displayed in our usage window?
Is the rate the same as what they’ve announced? (I’m slightly holding out that the pricing will be slightly less than what they are calculating now)
I am the one who has truly left Cursor, and it’s not about the prices, it’s about how much ■■■■ the team has been releasing over the last six months. It’s simply impossible to use.
They clearly said that yearly plan won’t change until it expires and is up for renewal. What more do you want?
Free plans aren’t guaranteed to stay the same for any amount of time. Especially not with a relatively new technology. Expecting them to is just silly.
This is confusing. Auto switches between models based on the task, but it has its own pricing. So sometimes you get Claude 4 Sonnet for cheaper than API rates, and sometimes you get GPT 4, and sometimes you get some internal model, all for the same price?
And once your months usage is exhausted, the ONLY option will be upgrade your subscription or increase your API usage spending cap.
I previously understood that Auto could select frontier models because like EC2 spot instances it was costing Cursor less as it was based on availability/demand. But now they are saying they “improved” Auto etc., which makes it sound like it is their own internal model. Perhaps it only fallback to another frontier model when their own infrastructure was not available to satisfy the request, rather than when 3rd party model infrastructure has spare capacity that they can get at a cheaper rate.
Very disappointing to not have ANY unlimited model to get by with when your included usage is exhausted. Considering that on Claude Code you just have to wait a few hours for the rate limit to reset, and GH Copilot has unlimited GPT4.1/4o for $10.
