[BUG] Usage limits bugged after double limits expired on new years.
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 a dedicated CC user and have brought many of my developer friends to the platform. I am on the MAX plan and have used it religiously since spring/early summar. With the MAX plan I was always able to do my work within the limits. In my entire history I have hit usage limits maybe 3 times. Since Jan 1st, I have been hititng rate limits with the same usage in an hour. My theory: In the Fall they changed usage limits and suddenly everyone was hitting them. A few days later they raised them to more reasonable limits. I have worked within those limits ever since. Then on Xmas-NewYears you gave us double usage limits, which was great. Unfortunately when that ended, the limits appear to have gone back to the limits when the change was first made in the fall that everyone hated instead of going back to the reasonable limits that were quickly adapted after the change. It's like the rate limits were reverted too far back in time.
Fall: Bad usage limits introduced
Couple days later: Reasonable usage limits applied
Xmas: Double usage limits until new years
NewYears: back to fall's bad usage limits
What Should Happen?
I should get the same reasonable usage limits that were in effect in december
Error Messages/Logs
"You have hit your usage limit"Steps to Reproduce
Use CC Normally, noticed rate limits hit far faster than in early december.
Claude Model
Not sure / Multiple models
Is this a regression?
Yes, this worked in a previous version
Last Working Version
_No response_
Claude Code Version
v2.0.76
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
Cursor
Additional Information
Fall: Bad usage limits
Fall/Winter: Reasonable usage limits
Xmas: Double usage limits
NewYears: back to bad usage limits. Note, this is not just a "feeling" of moving back down from double. It is noticeably worse than pre-xmas, like it was with the change in the early fall.
Please go back to the reasonable limits.
40 Comments
Had the same issue - MAX x20 user - same workflows hit the week limit in 3 days - never hit it before. I was not working over the christmas period - so I have not gotten used to the 2x promotion - I am doing the same workflow as I have for the last few months.
This is the first week I've ever hit my weekly limit on the MAX 20 account. Something doesn't feel right.
I had a brief convo with Claude: no Claude Code, no file uploads, no heavy compute; and it got me up to 18% usage with Claude admitting this is absurd: "Still absurd. A dozen text exchanges with web searches shouldn't be 17% of any paid plan's weekly limit.
Pro users were promised roughly 40-80 hours of Sonnet per week. This conversation has taken maybe 10 minutes."
<img width="1120" height="436" alt="Image" src="https://github.com/user-attachments/assets/b5e728f5-e061-4a56-9b58-8d26d9b94bdd" />
Same here. I was using Max plan (5X) before and never got my limits reached. Then suddenly after 31st, I got my limit reached in 2 days. Moved to 20X plan, and this is also hitting in 4 days. This is not good at all. Was using Opus 4.5 before, and same now.
<img width="1037" height="437" alt="Image" src="https://github.com/user-attachments/assets/b021c0e1-0f29-4637-aa15-f5f09dfd8570" />
I'm also getting this. Never came close to hitting useage limits before the new year - now unusable
same issue here!!!!
Same here!
I hit my weekly limit on my X20 plan within 3 days, normally only hit it around 5-6 days. Couldn't believe it, same project, indexed project that I've used for months, was shocked to see after holiday bonus went from x2 extra usage to /4 (standard). That's my guess either way. So I've gone ahead and canceled my subscription which is due for renewal in a few days. Hopefully they fix this issue.
How do we know if anyone from @anthropics is looking into this? Last time there was a usage bug it was fixed in a couple of days. We're now on day 6.
Just done a test with a Pro account, used 100% of session limit within 2 opus 4.5 prompts, nothing crazy. Reached within 10 minutes of the session starting. Definitely something wrong with the token calculations after reverting the 2x Christmas gift.
I'm having the same issue. x20 plan and I hit the limits within a few days. I added $100 of extra usage and I've burned through that in about 2 hours of my typical usage. I've been using CC on the x20 plan since last year and I've never burned through tokens this fast. I am using Opus 4.5 (which is amazing).
Sorry, feels like spamming here, but I've also tested an Enterprise account which I use daily for work, one prompt with Opus 4.5 took up 44% of my 5-hour session, crazy! It's never been that fast before, so 3 different accounts I've tested, x20, pro and enterprise (standard seat), all same issue.
yeah the last comment definitely makes me think they no longer are as generous with opus anymore and reduced it's usage to before they released 4.5. I was in the same boat, first time I ever hit a limit in 6mo of subscribing last week after the 2x was removed. But since I switched to
opusplanI haven't hit the limit again, making me suspect it's related to opus.Same with 20x plan.
same issue on 20x plan.
Also experiencing same on 20x plan.
also same, we are paying so much money and yet services are subpar, i think I will order codex now, not sure, if I will ever return to claude code
On the Pro plan and approx. 8 mins of Claude code using Sonnet 4.5 and I've burned through 30% of my quota.
same issue 20x
Now I can't see usage!? Wow...
<img width="1395" height="464" alt="Image" src="https://github.com/user-attachments/assets/5e9cbe0e-ee98-41dc-b9f7-ac58827f8ee9" />
Usage limits are removed while they look into this. I was at 100% opus usage before I saw this update, and now I can leverage Opus without issue. My reset wasn't until tomorrow morning.
Of course that happens while we Aussies are asleep :( Woke up and 30 mins later it was resolved. I was hopeful that they were fixing the issue, alas, seemingly not.
Has there been any official communication on this?
I'm certain my usage limits were impacted, as I neared my weekly threshold on day 2 of my usual workflow.
Further, my surprised inquiry w/ Claude into "what just happened here?" cost me _4%_ of my weekly limit.
That's absurd, and Claude agreed.
Well all things considered, 5-hour session/weekly usage is still very much terrible, even after Anthropic "fixed" the issue.
Same issue
Seems like usage is back to a reasonable level? Yes, usage limits suck but I can understand the need.
I wish Anthropic was more transparent with how these things are handled... At least in this case it seemed big enough of an impact that they should've made a public statement to prevent people from burning through usage and possible paying for a bunch of extra usage (and how they're going to make things right, if that's the case).
I think they should make a tier for unlimited use, too.
<img width="947" height="433" alt="Image" src="https://github.com/user-attachments/assets/bd2acc78-a3cb-42a0-92ed-c3e56fb1b6ae" /> This is after 2X 5-hour sessions on a fresh weekly reset., so basically within 9 hour's, already 40% weekly limit used and it only reset 12 hours ago... X20 Plan.
Experiencing the same. Big hit on productivity
<img width="993" height="422" alt="Image" src="https://github.com/user-attachments/assets/a3f5a381-ee78-41d2-b90d-7b432bcba76d" /> X20 Plan, reset 24 hours ago.....
Same here. Claude Max 5x, in use daily from 9-6 for the past 4 months, never hit a rate limit. Have hit it 3 times in the past 2 days, not even generating code. It's most definitely not the same allowance as pre xmas.
I checked my usage using CCusage (great tool!). And this is a perfect example. The week of Christmas, when there were double rate limits, I was able to use 400 million tokens and still have some usage left over. This week I have used 78 million tokens and I'm at 81% usage. Half of 400 million should be 200 million and yet I'm close to being maxed out at only 78 million. The math doesn't add up, usage limits are currently broken.
Environment Info
I've just ran into this too and looking at CCusage it looks like the cause it that while previous to today, tasks were being done on a number of different models (haiku and sonnet) this morning everything was using Opus 4.5 (as it looks like an update made this the default) which has a lot lower limits.
Very annoying as I was happy with the outputs from the Haiku and Sonnet and never needed to rely on opus before.
I totally agree, the limit is reached with very little actions performed
Same here I'm seeing a single request just consumed 16% of my session limit (pro user).
The requested resulted in just 10-12 lines of code being changed.
Granted it had some context in the thread history before the request, but I've never hit the limits so frequently like this before.
I just subscribed to Claude Pro at the start of the new year, previously subscribed to ChatGPT Plus. I thought to myself "The usage limit couldn't be that bad" 3 weeks in, I am considering cancelling and going back to ChatGPT. I love Claude so much more, it's smarter, both with writing and coding - which is mainly what I use AI for, helping me brainstorm and understand - but these usage rates are INSANE.
Anthropic, I hope you are reading this; If you don't raise the usage limit, you'll slowly start losing PAYING customers, and you'll notice the numbers start to decline. If you truly are the ethics driven company you claim to be, you'll raise the usage, or do something to make Pro and Max worth it, because for how much we paying, and how little we are getting, it makes Claude borderline unethical.
I am a new Claude Pro subscriber, having moved over from OpenAI and Gemini specifically to use Claude Code.
However, the current usage limits are rendering the tool unusable for professional workflows. I am hitting the daily usage limit after only 3-5 prompts in plan/edit mode. Furthermore, within 48 hours of account creation, I hit a weekly limit that has locked me out of both the terminal and the browser interface for days.
This consumption rate (3 prompts = 5-hour lockout) seems disproportionate and makes it impossible to finish a task. I am building 2 webapp to take to market and quite honestly I can't even do 30min of work and the limit is reached.
I have noticed similar reports in the community regarding usage tracking discrepancies:
Issue #18786, #18381, #18349, #22435, (GitHub*), and many many many more spanning from Jan 1st to today... How is that possible? Is @anthropics ignoring us?
I use the MAX subscription and also noticed that it quickly hits its limits, even though I only used it for minor fixes on my project. Previously, the subscription lasted much longer.
Yep, I'm noticing same behavior.
Same here. The problem is not the models (even though in some Anthropic reports they claim that Opus could consume less because it reaches the solution faster...), it is about the completely opaque way to handle limits.
Just taking measurements in the usage progress bar and checking out the consumption of the tasks... there is a nonlinear pattern. Pro plan was more than enough to experiment and do a lot in a non-production environment, but with the actual limits you cannot even do that.
My main problem is the weekly limits, because I never reached it before this week. Now, by my calculations, I will always reach it by Wednesday... what makes the Pro plan useless in most cases.
Regarding the session limits, I notice something weird, but the effect is more subtle, and I switched to Opus and the comparison is not fair until I get more experience with it.
Same for me, I bought an Pro year plan and hit the limit with 1 question... I asked a refund.