[BUG] Claude hitting usage limits in no time, something broken for sure
Preflight Checklist
- [x] I have searched existing issues and this hasn't been reported yet
- [x] This is a single bug report (please file separate reports for different bugs)
- [x] I am using the latest version of Claude Code
What's Wrong?
I am facing limit issues and I believe there is certainly a bug in your new releases since I did not face this issue at all in my last 4 months of claude usage. Since last 3 days, my limit is getting used very quickly. Yesterday, it got over at 530 PM and I had to wait for 3 hours. Then it got active at 830 PM again and again went off within 2 hours. I had to stop my work at 1030 PM yday. I started again at 730 AM today and again it went off in < 2 hours. So I added 5$ to check is the issue at my end or at your end. (In confusion, 5$ got added twice). To my shock, the 5$ just vanished in just 1 min, without me doing anything serious at all. I just gave one instruction and 6.28$ spent.
What Should Happen?
Claude limits logic should work as it has been working earlier. There is some serious issue at your end since I have not changed my claude usage or my coding style at all.
Error Messages/Logs
Steps to Reproduce
Launch claude session, do some work and limits gets over in no time.
Claude Model
Sonnet (default)
Is this a regression?
Yes, this worked in a previous version
Last Working Version
No response
Claude Code Version
2.1.2 (Claude Code)
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
Terminal.app (macOS)
Additional Information
I can provide any additional information if needed.
Bug Report: 10x Token Consumption Increase - Max x20 Subscription
Environment:
- Claude Code version: 2.1.2
- Platform: WSL2 (Ubuntu) on Windows
- Model: Claude Opus 4.5
- Subscription: Claude Max x20
Description
After 3 months of consistent usage, token consumption suddenly spiked. Very noticeable since January 8th.
Before: Max out weekly x20 quota over 50-80 hours of work. Rarely hit 50% session usage.
Now:
- Went from 0% to 23% quota in under 5 hours of trivial work
- Fixing two trivial test failures + running tests: 20% → 23% in ~30 minutes
- That same 30 minutes used 32% of session budget (I almost never hit 50% session usage)
- My Max x20 now feels like a normal Pro subscription
Key Point
I made no changes to my configuration or workflow. Same MCP servers, same setup, same type of work - just file moves, small refactors, and running test suites.
Impact
Unusable for professional work at current consumption rates.
Related Issues
Same issue as #16951 and #17016.
edit: Also reported at https://github.com/anthropics/claude-code/issues/16157, as this one seems like a duplicate
I faced the same problem as reported, though i am inn 2.1.1 i notice this in last few days.
+1
Fixed for me by going to v2.0.64 for temporary fixing. Probably don't need to go that far back but went off someone else's recommendation. Seems this is working for some but not all. Hope this is fixed soon.
Fixed for me by going to v2.0.64 for temporary fixing. Probably don't need to go that far back but went off someone else's recommendation. Seems this is working for some but not all. Hope this is fixed soon.
Thanks for pointing this out, I will downgrade immediately.
I am scared to even launch CC now and wondering what could be done to stop it eating my tokens like a blue whale!!!
I have the same issue on pro plan, i hit the session usage in no time , even the weekly limit is at 42% and it resetted yesterday at 16:00, i never come close to 50% for the whole week. I will downgrade until it gets fixed.
Issue: Rapid usage drain in Claude Code v2.1.2.
After upgrading to v2.1.2, my Pro account usage climbed to 82% in 30 minutes, despite the tool being idle. I had to log out of the terminal and revoke the access token in the Web UI to stop the consumption. This issue was not present in the previous version. Note: The VS Code extension remains unaffected and works as expected.
Just a suggestion. If you actually want anything done about a potentially big problem, provide DATA. I read these kinds of posts to see if it is safe to upgrade to the latest release, and I see a lot of these posts and ALMOST NONE OF YOU actually dig into your actual usage numbers to PROVE "odd" usage. Do you expect Anthropic to come to your basement and dig through your dirty laundry to figure out the problem? How many of you are actual programmers that have any clue how Github issues NORMALLY work? Because I'm not seeing anything actionable in ANY of your reports. Makes me feel sorry for Anthropics devs.