– We never signed up for any credits or tokens — we all came for the service of “500 requests and an unlimited pool.”
– We never agreed to a pay-as-you-go API model.
– We never agreed to new pricing.
You’re the ones getting credits from investors — and from us, a credit of trust. But now you’re failing to deliver and not fulfilling your responsibilities.
The only thing you’ve done consistently is make things worse. What exactly have you improved?
– You restricted free DeepSeek models to 60K context.
– You gutted and killed the free plan — there are no premium models there.
– You reduced the 500 requests to 225, and then downgraded it to a paid API where you can barely make a few dozen calls.
Now you’re proud of having unlimited ‘TAB’ and ‘AUTO’, but there’s nothing to be proud of — those are just crumbs left from the pie.
We’re not a cash cow for you to treat us this way.
I has used Cursor for over a year, Cursor used to be amazing. It could read the full context, and the requests never ran out even if I had to wait 1–2 minutes.
So moving to the new Pro plan will get a lot of complaints. The usage-based pricing went up from $0.04 to about $1 per request (Slow req → Token req).
I think the plan would be better to just give us more token requests and raise the price to $25–28 (The usage-based system used up $5 of mine in a single prompt, so please don’t suggest that I turn it on.). That would still be okay for us and make Cursor easier to compete with others.
They’re actively cutting back functionality because they can’t afford to maintain it.
It looks more like a drawn-out bankruptcy phase, where they keep changing the rules just to stay afloat.
They used to say, God forbid we ever raise prices — it would hurt our users.
And now? Are we just supposed to swallow it?
It reminds me of what happens with companies that are on the verge of collapse — they desperately try to squeeze out as much money as possible before the end. It’s a clear pattern.
Thanks for the responses. Unfortunately, what it reveals is that in Auto mode, it’s really up to somebody else to decide which model we get, so despite having success with 3.7, I’ll be downgraded for unspecified reasons.
I was under the impression that beyond the $20, responses would be slower, but instead, the quality is a lot lower, and you’re okay with that.
Okay, but how do I plan a month’s worth of work when I’m unsure about how the system charges my account or which models are used in Auto mode? Previous descriptions had it as each user prompt was a request and worth $0.04. Now, the new pricing model implies every agent action is also $0.04. That will eat up the credits!
I have purchased for one year the subscription with the 500 fast requests and the rest slow. Now it tells me I have only $20 for the usage. That was not what I paid for and also I do not have the option to go back to the previous version of the plan. I need a solution.
Here’s what happened when I contacted Cursor support about the pricing changes:
Timeline:
July 4, 2025, 22:21: Emailed [email protected] about extended rate limit issues
Initial response: Generic AI assistant response with standard troubleshooting
My follow-up: Detailed explanation of the actual problem
Final support response: “You’re right - I’ll connect you with a teammate who can better investigate this extended rate limit duration. They’ll get back to you as soon as possible to help determine why this is lasting longer than expected.”
July 9, 2025: Unsubscribed before new billing cycle due to lack of follow-up
Today (July 12, 2025, 01:28): Still no response - 8 days later
The Issue: I was locked out of Claude 4 Sonnet for 71+ hours with no clear reset timeframe, which contradicts your documentation stating limits reset “every few hours.”
The Result: Your support team’s failure to follow through on their promise led directly to me cancelling my Pro subscription. I unsubscribed on July 9th because waiting 5+ days for a promised investigation with zero communication was unacceptable.
The Problem: This isn’t just about pricing transparency - it’s about basic customer service. When you specifically promise a human will investigate and provide answers “as soon as possible,” then go silent for over a week, you’re actively driving away paying customers.
If you’re consolidating feedback here, please include this: broken promises and abandoned support tickets are causing subscription cancellations.
As you send messages in Cursor, you will consume the credit at the same rate you would via the LLMs direct APIs.
This doesn’t make any sense. Why when we pay up front do we pay the “retail” rate. The reason I pre-purchase a plan at DisneyLand is because it is cheaper than purchasing the food when I get there. That is the whole point of pre-purchasing something. Not to mention that the volume alone should give Cursor the ability to bulk purchase tokens at a steep discount. The cost/benefit just isn’t there with the new pricing.
1、The new pricing model is like 4K video communication — you can only use it about 5 times a day, after which you’re locked out for several days. And the model quality isn’t even that good.
2、The old pricing model is like 2K video communication — you can use it daily, and its model quality is about the same as the new one.
3、“Auto” is like an ancient telegraph or Morse code — it’s unlimited, but highly prone to errors and random behavior.
History is eerily similar every time.
At first, their contract promised 500 fast and unlimited slow requests. And I didn’t speak up — because I could still use slow requests.
Then they introduced the $0.04 charges. And I didn’t speak up — because I didn’t use the $0.04 option.
Then they silently rewrote the contract into a new pricing model. And I didn’t speak up — because it still worked for a few days.
Then they manually deleted my post questioning the pricing. And I didn’t speak up — because I wasn’t the only one they silenced.
Then they used announcements to deceive users, saying emailing ([email protected]) would restore the old plan. And by that time no one was left to speak up.
I saw your recent pricing changes for Pro plan and I wanted to clarify a few things. So I’d purchased a cursor pro plan on student discount for 1 year during May 2025, now I’ve been using cursor after this new pricing but according to new pricing we can use frontier models on API pricing upto $20. But I saw a few posts on X and Cursor forum that the new pricing will be applied in the new billing cycle so I wanted to clarify whether it’s true or not.
currently it seems Cursor becomes crazy in charging.
Is it possible to give another tier of pro in 50 dollar per month?
it is only 2nd week in July, usage charge was higher than 50 dollar, above the basic subscription fee.
Another problem.
without max mode, it would delete 1/4 of the code for adding one number to choice.
But with max mode, it charge multiple times per minute.
does this mean we could only use cursor for code less than 2000 lines?
maybe I should try my best to learn linux for Claude Code, or I shall use Gemini Cli instead… Too painful now, even though it is claimable but unrealistic.
There value was their client base, they’ve alienated them. Other than that they’re a rebranded vscode fork. 9bn valuation is hilarious and just shows the insanity of VC’s.
Someone asked in another thread, but I can not find it anymore, and since it is very related to this topic, I will ask it here again:
What is currently Cursor’s value proposition as a product?
Correct me if I am wrong, but Cursor happens to be in this moment just a proxy to different LLM APIs, for the same price you could use those APIs if you use them directly with VS Code fork on top of it. Which is where it makes me think: what’s the point of subscription at all? When I subscribe on something I do it to save money and get it below the price. If I need to pay 10 McDonnalds meals in advance for the same price they would cost me when I get there, that’s not subscribing, that’s crediting McDonnalds.
Without going too much into what makes or doesn’t make sense for Cursor as a company, I am seriously curious what we are buying here, and if there is any vision for Cursor beside pump the valuation and potentially get bought out by Anthropic, Google or Microsoft?
What have they actually done?
All they did was fork VSCode with a nice UI, take an open-source LLM, tweak it slightly, and call it “TAB” along with a few weak “free” models — and that’s it.
They haven’t built anything substantial. They’re just a market middleman — repackaging and reselling under a polished facade.
What they should have done:
If they were smart, they would have taken a powerful open-source model like DeepSeek, made it the core of the product, and hosted it on their own (or rented) servers.
But their goal isn’t to build a great product — it’s to squeeze profit. This is a student-led operation focused on monetization, not innovation.
What they’re doing now:
Cutting back free models.
Promoting overpriced premium ones.
Completely missing the point of sustainable product design.
What they don’t understand:
You can build a free product that’s actually usable for real work — with no crippling limits — and then monetize with optional paid features or subscriptions.
If the core product is useful and reliable, people will gladly pay for extras.
That’s how you build trust — and a lasting business.
Cursor give us the “option” to send an email to hi@cursor…
But almost 3 days later without response, they still waiting my monthly subscription finish. The only model really usefull is claude, the others options is totally dumb in MY USE. So why shouldn’t i not cancel cursor and go to claude code? I started with you guys in january, but right now cursor is for “vibers” who use 10 minuts each day.
It’s particularly ironic that Cursor hired Boris Cherny (the ‘soul’ of Claude Code) and Cat Wu (Claude Code PM) from Anthropic, yet the product they originally built still provides 29x better value than their current employer’s offering.
