claude-code - 💡(How to fix) Fix [BUG] Weekly Limits – All Models [1 comments, 2 participants]

Preflight Checklist

• I have searched existing issues and this hasn't been reported yet
• This is a single bug report (please file separate reports for different bugs)
• I am using the latest version of Claude Code

What's Wrong?

As soon as the "Weekly Limits – All Models" are completely exhausted, I completely lose the ability to use Claude, even though the Sonnet usage indicator still shows less than half of the resources used. In this situation, I can no longer access Sonnet, the situation with Opus 4.7 is strange, the model consumes tokens extremely quickly. At the same time, Sonnet 4.6 completes many more tasks while using significantly fewer tokens.

This behaviour seems counterintuitive and unfair. It forces me to burn through tokens unnecessarily with Opus when a more efficient model — with remaining capacity — is sitting idle. I seriously doubt this was Anthropic’s intended design, but if it was, I believe it urgently needs to be corrected.

A couple of improvements would make a huge difference:

Allow access to Sonnet 4.6 even after the “All models” pool is exhausted, as long as the Sonnet-specific quota has not been used up.

If both models draw from a shared token pool, make the allocation smarter. For instance, use more of Sonnet by default and limit how disproportionately Opus 4.7 can drain the shared budget. A system that dynamically shifts load towards Sonnet would be far more efficient and user-friendly.

One way or another, you will come to solve this problem, it seems to me, because this is ABNORMAL

I am a user of Web Max Plan X5

What Should Happen?

lower token consumption in Opus 4.7 redistribution of load between Sonnet 4.6, right now they are both dependent on 'All Models' so they are required to redistribute the load, I am losing the ability to use it in the literal sense of the word, I am losing my money waiting for the limit reset due to the malfunction of Opus 4.7

Error Messages/Logs

-

Steps to Reproduce

work in these models

Claude Model

Opus

Is this a regression?

No, this never worked

Last Working Version

Claude Code Version

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Other

Additional Information