Feedback: New Pricing Model

While I’m excited to hear about the new pricing model, it’s missing some key components that I hope you’ll address soon.

1. Transparency Issues

The new pricing model lacks the transparency of the old system. Since the change has been live for a a-bit now, the lack of documentation is disappointing.

We need clear information about:

  • How you determine when to rate-limit users
  • How long rate-limiting lasts
  • How different models impact rate-limiting decisions

2. Value Proposition Concerns

A key reason I brought my project to Cursor was the unlimited slow requests feature. Removing this significantly reduces the product’s overall value.

My situation: Earlier in my project, I used Claude Code exclusively, spending thousands per month. With your new pricing model, my options as an independent developer are limited to either:

  • Stop using the service entirely
  • Drastically reduce usage

Even though I didn’t heavily use slow requests in the previous model, my costs still felt manageable at $100-200/month versus the thousands I was spending before.

3. Suspected Cost Increases

This may be speculation, but it strongly feels like pricing has been adjusted upward for the new plans. My month-over-month analysis of the same model shows increased costs for similar requests.

Final Thoughts

I understand you need to make changes to stay competitive—other AI firms have rolled out similar high-cost plans. However, for independent developers without enterprise budgets, this new system desperately needs better transparency.

Please update your documentation and supporting processes to clearly describe:

  • How you calculate billing
  • When charges are applied
  • How requests and tools contribute to usage costs

Until these transparency issues are addressed, I’m opting out.

Edit:
Claude 3.7 Extra requests vs Claude 4 extra requests x5 increase with new plan. Obviously not a 1:1, rates went up, part of the Screencap requests were on the old system, etc..

13 Likes

incredible feedback, this is how I feel as well.

3 Likes

hi @NoOne_Cool and welcome to the Cursor Forum.

Sorry, it seems that so far we havent gotten to your post as there are duplicate posts on the same topic. Thank you for the constructive feedback.

Cursor Team has already responded to those questions in other threads and made some adjustments based on feedback in specific cases.

I will try here a rehash but its not the final or full info:

  • The plan determines the rate limits. Higher plans have higher limits and are intended for users who also on old plan needed more heavy usage.
  • The specific info on rate limits and duration has not been released by Cursor Team but they have clarified that most users do not reach limits easily and it does allow much more requests to be used than with the legacy plan.

What impacts rate limits = resource usage:

  • Heavier models require more resources and therefore increase usage
  • Larger context requires more resources and therefore increases usage (this may be the case for you as you use Sonnet)

From my own usage of my personal Pro plan, I have not yet reached the rate limits with daily usage of 10 to 20 000 lines of code edited by Sonnet 4. However I take care of context size and dont add unnecessary context.

There are two limits:

  • Burst limit: allows intensive usage of a model in short period but refills slow
  • Local rate limit: allows continuous regular usage of a model through time and refills regularly.

In case you know you will have heavy usage there are options:

  • Usage based pricing: allows you to continue over the rate limit and is charged on resource usage (model and context).
  • Pro+ and Ultra plans: allow for 3x or 20x more usage.

The current Pro plan has not been changed in price. The larger usage plans are only for those who know they will have much more usage and do allow them higher limits before reaching usage based pricing.

You can see the breakdown of current usage when you mouse over the cost, it will show how many tokens were used and for what. I recommend looking into context sizes and if you can remove unnecessary parts to get more out of the current plan.

Theres a ton of threads regarding the new price so im going to piggyback off of this and hope I get an answer.

How is the price of each model determined now?
Previously it was as simple as ‘$x per prompt’ which it looks like now but its deceptive as its now.

claude-4-sonnet for example, shows 1x token, however I get prompts now that are $0.67 per prompt!
Previously this would hit 25 tool calls and ask for a resume, and I would know its 1 more ‘cost’

Does changing to legacy change this also? Seems like im getting ripped off now.

1 Like

Requests above included usage in a plan are now charged based on API price of the model (e.g. the number of tokens consumed by the request, multiplied by the API cost per token of the chosen model).

When hitting a rate limit, you can switch to the ‘auto’ model for unlimited use, enable usage based pricing for charges based on API rates, or upgrade the plan tier to get more included usage and higher rate limits.

Cheers, thats increased the cost by a crazy amount. Might shop around and see what else is around :slight_smile:
Hopefully something as good or close.
Unfortunately auto just gives bad results 99% of the time.

1 Like

Interestingly I went back and looked at my history.
April - 5.2k requests - $40 over pro plan
This month so far - 1.7k requests - Somewhere above $40 and below $50 above plan

3 Likes

I’m experiencing this as well - paying more for less…

1 Like

Sorry I can’t help but I just came here to say the same thing… I was merrily just using Claude 4 in non-max mode then suddenly get a message saying they are charging me $60. As you say, some prompts are getting charged like they are MAX or something.
Yet the pricing page still just says it’s 1 request and requests are $0.04.

If I could have the slightest clue as to when and why it got charged like that then maybe it would be OK, but for now I’ve opted out. (found that option in the Advance settings in my profile)


Opus really tired today, not starting, thinking for minutes, doing 2 lines of changes then sleeping, errors on edit_tool, really weird situation we have arrived at.

1 Like

@vibe-qa please file a detailed Bug report about the issues you’re having, so they can be investigated and cause found out. This is not expected behavior without bug reports your specific edge case can’t be fixed as it’s not happening for most other users. You can tag me with @ in your report so I don’t miss it.

There are plenty of bug reports with the same behavior, not going to spam about it.

Ok but the screenshot you posted gives me no info about what issues you are facing with Opus. I would like for the issues to be checked and resolved. If you can otherwise point me to a relevant thread that would help as well. Errored, Not Charged has many valid cases and causes.

have the same feedback here, was using cursor for pretty intensive task,only with claude 4 and was pretty happy with how it was working.
It was clear I knew the number of request I had left and how much I will pay.

I’ve been switched to the new plan but didn’t got notified. By luck I went to look at my usage and saw that it was consuming my money fast. No alert, nothing …

I am happy to pay for the value, but I need to know what I will pay.
Right now it’s unusable for me …

1 Like

Docs are being updated to clearer explain the cost incurred per request.

You can also mouse over the request charged to see what the tokens usage is (input/output cache read/write) which is how API providers count resource usage per model. The usage based pricing depends on model used and tokens used. But that will become clearer again with updated docs.

You cant see this on non paid for requests though which makes working towards more efficient requests much harder.

1 Like

@Reznal Yes I agree, feedback has been forwarded to Cursor Team already.

I like the new pricing model. I think per-request pricing models are just unfair to both the company and the users. Users who don’t use large contexts should be able to chat more than users who do, and that’s how upstream billing works. It’s nice to see better MAX mode integration with the plan too. It always felt odd that there was no way to exchange my normal requests for MAX requests. I wish the transparency were better and the docs were updated before the rollout because it’s just unnecessary frustration for what is otherwise a positive change. PS. I was also about to mention this cringeworthy “unlimited requests” line on the pricing page, but it seems like you have already changed it :slight_smile:.

1 Like

I don’t understand why we are automatically switching from “Usage Based” to “Included in Pro”, back and forth, without even knowing what’s happening behind the scenes.

When usage based pricing is on, once you reach your plans limit it will use usage based pricing, similar to before.

Here is a short intro into recent plan updates:

  • Plans switched from monthly 500 fast requests to no monthly request limit plan with usage based rate limit depending on model and token usage.
  • There is a burst limit that allows intensive short period usage but refills slow and a local limit that refills regularly but with smaller amounts.
  • Once rate limit is reached usage based pricing takes over to continue uniterrupted usage, model power and token usage is charged at API pricing (+20% markup), the Usage records with $ should have on hover the exact token amounts used.