I'm being charged requests when "claude-4.6-opus-high-thinking" model errors out with no token usage

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

The model “claude-4.6-opus-high-thinking” is self-triggering but quickly errors out with no token usage but I’m still being charged requests.

Steps to Reproduce

By running an agent request using the “claude-4.6-opus-high-thinking” model.

Expected Behavior

If the model errors with no token usage, I should not be charged a request.

Screenshots / Screen Recordings

Operating System

Windows 10/11

Version Information

Version: 2.4.23 (system setup)
VSCode Version: 1.105.1
Commit: 379934e04d2b3290cf7aefa14560f942e4212920
Date: 2026-01-29T21:24:23.350Z
Build Type: Stable
Release Track: Default
Electron: 39.2.7
Chromium: 142.0.7444.235
Node.js: 22.21.1
V8: 14.2.231.21-electron.0
OS: Windows_NT x64 10.0.26200

For AI issues: which model did you use?

claude-4.6-opus-high-thinking

Does this stop you from using Cursor

Yes - Cursor is unusable

@deanrie If you could take a look at this issue, I’d really appreciate it. When I opened Cursor quickly it still auto-triggered the agent multiple times and charged requests.

same here

same issues

Hey, this is a known issue. Errored requests can show up as charged in the usage dashboard at first, but billing should auto-fix it and mark them as Errored, No Charge after some time, usually within an hour.

Check your dashboard a bit later and you should see it corrected.

A couple of things:

  1. You’re on version 2.4.23 from Jan 29, which is pretty old. Please update to the latest version. Older clients can handle errors differently and may hit errors more often.

  2. If the requests still show as charged after a while, email [email protected] and the team can review your account and fix the usage manually.

The team knows about this reconciliation delay. Your report and the confirmations from others in this thread help with prioritization.

1 Like

Are you aware these are being auto-triggered without me manually using the agent?

For me, it was constantly flipping between errors and reconnecting and sent so many requests :joy:

Version: 2.6.20 (user setup)
VSCode Version: 1.105.1
Commit: b29eb4ee5f9f6d1cb2afbc09070198d3ea6ad760
Date: 2026-03-17T01:50:02.404Z
Build Type: Stable
Release Track: Early Access
Electron: 39.8.1
Chromium: 142.0.7444.265
Node.js: 22.22.1
V8: 14.2.231.22-electron.0
OS: Windows_NT x64 10.0.22631

For the version upgrade recommendation, I’m still using this older version because of the issue here: Terminal problems when used by agent - #27 by j-173

I see the note about auto trigger, that’s an important point. The problem isn’t only that errored requests get billed, it’s also that the agent can loop and generate requests without you doing anything. I flagged this to the team separately.

@Liu_Zhiliu, thanks for the version info. I can reproduce it on 2.6.20 too, so it’s not just an old client issue.

@j-173, got it about the terminal bug that’s blocking you from updating. For the auto trigger issue, can you send a Request ID from one of those auto requests? You can copy it via right top corner of the chat > Copy Request ID. That’ll help us see what’s triggering the repeated requests.

On billing, the team is aware. If after reconciliation the requests still show as charged, email [email protected] and they can fix it manually.

Thank you for the quick replies @deanrie.

I already deleted the chat sessions for those auto requests. If there’s a method to recover, I’m happy to provide the Request ID’s.

Hey @j-173, @k_martin, and @Liu_Zhiliu!

The usage accounting issue for errored requests has been fixed in a recent Cursor update. Updating to the latest version should resolve this. If you’re still experiencing this on the latest version of Cursor, please raise a new thread and we’ll take a look.

1 Like