Cursor Costs Are Climbing Without a Clear Reason

Ideas Feedback
26

Hey @kfkhalili, thanks for the honest feedback. It sucks to hear Ultra didn’t work out.

We already fixed part of the 3.0.x issue in 3.1.x, where one prompt could turn into dozens of billable requests. Details are in post #18. But if your case was more like the needle barely moves compared to competitors for the same amount of work, that’s about overall token efficiency, and it’s a signal the team is paying attention to.

If you decide to give it another shot and still see quota burning out of proportion, please share the Request ID and the version here and we’ll check what’s happening on the server side. I can’t give an ETA on bigger token efficiency changes yet.

27

Hi @deanrie, thanks for writing back.

I used Cursor for more than a year and recommended it to everyone. I don’t know what you guys did but the burn rate went completely crazy towards the end of last month. You say issues were fixed and I can appreciate that, but how was I, the user, compensated? Was I given more credits? Or had some limits reset? I don’t think so.

Anyway I am not likely to spend $200 (plus VAT) just to watch it all burn in 4-5 days, and even for requests that failed or didn’t return a response. You’re sorry and I’m sorry and we’re all sorry. Good luck with your future endeavors.

30

My company has a team license for Cursor, but we also have one for Claude for non-dev use. I use Claude Desktop quite a lot to help other teams create skills and projects, and lots of random questions I don’t want to burn Cursor credits for.

When my Cursor usage dried up much earlier than usual I started testing out Claude Code in the meantime… While I still prefer Cursor’s product, the difference in efficiency is night and day. I’ve been using CC for all my dev work for the last two weeks (in addition to lots of non-dev usage) and haven’t hit a single limit yet.

31

Thanks for everyone chiming in hopefully we can provide information to the cursor team and get this matter resolved. With enough people noticing this, I think its time for this matter to be really escalated, properly do root cause analysis on, provide a fix and explanation to customers for transparency.

32

Bump. Somethings broken here. (Or perhaps by design for more revenue?)

Our bill is thru the roof this month.

Same workflow as always. Still mostly using claude models.

But our bill is up almost 10x this month

34

Hey @dom1, thanks for the report. On the by design point, no, we’re not intentionally making anything more expensive. We fixed part of the issue in 3.1.x from 3.0.x, where a single prompt could turn into dozens of billable requests details are in post #18. Claude models are no longer auto-marked as Max, and in typical cases one prompt is counted as one request again.

A 10x jump on the same workflow is definitely worth checking. To understand what’s happening on your side, can you share:

  1. Your Cursor version Help > About. If you’re still on 3.0.x, updating alone can noticeably reduce usage.
  2. The Request ID for one of the expensive prompts. In the chat, open the top-right menu > Copy Request ID. With that, we can see what actually happened on the server side.
  3. Which models you’re using and whether Max mode is enabled anywhere.

For your specific bill breakdown why this cycle ended up 10x higher, it’s best to email the team at [email protected] since they can see account-specific details.

Once you send the Request ID, I’ll check whether it matches something we’re already tracking or if it’s a separate issue.