[BUG] Opus 5 xhigh/max usage suddenly drains 0→100% within seconds across multiple accounts

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?

What's Wrong?

Since today, September 18, 2026, I've been seeing extremely unusual usage consumption in Claude Code when using Opus 5 with xhigh/max effort.

I started using Claude Code at around 08:00 AM Romania time (EEST / UTC+3) and everything behaved normally for the first few hours.

At around 11:00–11:30 AM EEST, something suddenly changed. From that point on, my available usage started disappearing extremely fast.

I tested this across around four separate Claude accounts, and the same thing happened on all of them. In some cases, an account went from almost no usage to effectively 100% used within roughly 10 seconds of starting a request.

Across all four accounts, I managed to exhaust nearly all of the available usage in roughly 10 minutes total.

I know that xhigh/max uses significantly more tokens than lower effort levels. I've used these settings many times before, including for large coding and agentic tasks, so I'm familiar with the higher consumption.

What I'm reporting is not simply that xhigh/max is expensive. The problem is that the consumption rate suddenly became completely different from what I had been seeing before, including earlier that same morning.

Timeline

Date: September 18, 2026
Timezone: Romania — EEST / UTC+3

Around 08:00 — started using Claude Code; usage looked normal.
08:00–11:00 — normal usage behavior.
Around 11:00–11:30 — unusually rapid usage consumption started.
After 11:30 — reproduced the same behavior across multiple accounts.
In total, around four accounts were effectively exhausted within roughly 10 minutes.

Usage was depleted at an extreme rate.

The most severe case I observed was roughly:

0% used → 100% used in about 10 seconds.

The fact that this started suddenly and happened across several different accounts during the same time window is why I believe this is worth investigating as a possible usage-accounting or backend issue.

What Should Happen?

I expect xhigh/max to consume substantially more usage than medium or high effort, but the usage should remain reasonably consistent with the actual amount of work being performed and with previous sessions using the same configuration. It should not suddenly consume an entire usage allowance within seconds under comparable workloads.

Error Messages/Logs

Steps to Reproduce

Start Claude Code CLI.
Select Opus 5.
Use xhigh/max effort.
Start a normal coding/agentic task.
Check usage before and shortly after starting the request.

In my case, usage started dropping almost immediately and at a rate I had never seen before.

Claude Model

_No response_

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.276 (Claude Code)

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

PowerShell

Additional Information

_No response_