What is $20 usage when the model refuses to continue at $16.31? Forcing me to upgrade to Pro+

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

I decided to try claude opus 4.1 and use the $20 of included usage to see what it could do. Everything was great until it reached $16.31. Then I get a popup saying choose Pro+ for $60/month.

Why did it stop at $16.31. Even if it had stopped at $20, I would have enabled my spend limit to allow overage. This made no difference when I did enabling thinking that the IDE believes I’m at $20 usage limit.

I am curious why the IDE is prompting me to change plan and not simply just charging me overages.

Steps to Reproduce

Used claude opus 4.1 specifically and not in Auto mode.

Expected Behavior

Keep using the model until $20 usage and then allow me to continue if I set a spend limit.

Screenshots / Screen Recordings

Operating System

Linux

Current Cursor Version (Menu → About Cursor → Copy)

Version: 1.5.5
VSCode Version: 1.99.3
Commit: 823f58d4f60b795a6aefb9955933f3a2f0331d70
Date: 2025-08-25T17:40:25.290Z
Electron: 34.5.8
Chromium: 132.0.6834.210
Node.js: 20.19.1
V8: 13.2.152.41-electron.0
OS: Linux x64 6.8.0-78-generic

Additional Information

In the end I switched to Auto as it seemed the only way I could continue. This happened mid-feature development. This is quite a confusing experience. I feel I’m $4 shy of usage and being forced to upgrade/change my plan instead of using spend limit. This is another example of getting ~70% the way through a major feature and hitting a brickwall which seems characteristic of my Cursor experience when trying to use it for complex jobs.

Does this stop you from using Cursor

No - Cursor works, but with this issue

1 Like

Hey, the reason this stops so early is because the system ensures you have enough remaining usage for the response the model you’ve chosen is about to take.

Claude Opus is an extremely expensive model, and Cursor will intelligently estimate the final cost of your next request based on the model you’ve chosen, and the context that’s about to be submitted to the model.

You should be able to switch to a cheaper model, like Sonnet, and continue to use your remaining usage, or as prompted, you can upgrade or change your usage limit to allow you to continue with Opus, if you choose to do so!

What kind of crazy large prompt would take 4$+ anyway, even with opus? You should probably dial back your context or open a new chat. And use cheaper models. Also Cursor usually provides at least around 10$ worth of bonus credits on 20$ plan, but I guess that does not count towards this estimate.

Using Opus as your primary model on the $20 plan is not what the Pro is intended for. You’d only get a handful of requests. Also grok-code is free right now but even outside the promotional period it is much cheaper than most models and still very capable, so that can help stretch your $20. Gpt-5-mini is also a step above Auto and free/cheap.

It would not take much to hit a $4 prompt with Opus.

Understood. Can you clarify whether I should be able to simply increase my spend limit on Pro and use Opus beyond $20 included?

I could not, and only got the IDE prompt to tell me to pick Auto or upgrade to $60/month.

Im aware of the increase in costs and got quite a bit done on $20. I wanted to finish at maybe another $10. But it would not allow me to use Opus with increased spend limit.

It feels like Opus is not available unless I subscribe to $60/month or there is a bug.

best way to learn currently is do a trial for a month and use auto . and then use the background agent in the brouser that is something like gpt 5 or something then have it auto save to git hub if you give it access. and Then make it make it into a prompt for you for the agent copy the code or info it gives you into your agent and make it make the front end ui for you to work with . use the background agent the backend stuff like the core stuff . also allding piecesos app to your machine and cursor would hep too . im currently working on how i can ask the pieces copilot for a solution and then once i ok it the agent implements it with out having to cut an paist it in .

That’s just how Opus works. Easily 2-4$ per task with documenting, reporting, debugging.

You can enable additional usage here. Have you tried that and see if that works with Opus?


It did not work for me. As I reported it just says I have to subscribe to $60 plan. Which is my main problem.

Interesting, so I wonder if certain models, like Opus, just can’t be used with usage-based pricing.

claudia opus is so expensive no one can afford it, so its pointless ha!

I’ve never used it. My only request to Opus would be: “Send me money but make sure you don’t get caught!”. Otherwise its probably not worth it.

This topic was automatically closed 22 days after the last reply. New replies are no longer allowed.