anthropics / anthropics/claude-code

[BUG] Weekly usage jumped from 0% to ~60% after two short conversations; "Approaching weekly limit" banner contradicts displayed 27%/13%

Open
#92,906 1 comment 2 reactions 0 assignees View on GitHub
area:cost area:tui duplicate
Dominant language
Python
Stars
145k
Forks
23.1k
PR merge metrics
PR metrics pending

Description

### Preflight Checklist

- [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) 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?

On Max (5x). My weekly usage period reset on Tuesday, September 8, 2026. Shortly after the reset I had **two short conversations** on Fable 5.1 — brief exchanges, no long-running or agentic tasks. My Fable usage indicator then jumped from **0% to approximately 60%**.

Later the same day, after running only three or four short tasks, I received an **"Approaching weekly usage limit"** banner. At that exact moment the usage panel showed:

| Metric | Value |
|---|---|
| All models | 27% used (resets Tue 7:00 AM) |
| Fable | 13% used (resets Tue 7:00 AM) |
| Current session | 49% used |

The banner and the displayed percentages contradict each other. The banner stated the limit resets on September 15, consistent with the period having started September 8 — so hitting the limit on day one of a fresh period does not match my actual usage.

The banner has since **disappeared on its own** with no change on my side.

### What Should Happen?

Usage should accumulate proportionally to actual token consumption, and the warning banner should be consistent with the displayed usage percentages.

### Error Messages/Logs

```shell

```

### Steps to Reproduce

Not reliably reproducible on demand. Observed sequence:

1. Weekly period resets (Tue, Sep 8, 7:00 AM)
2. Two short conversations on Fable 5.1
3. Fable usage indicator jumps from 0% to ~60%
4. Three to four additional short tasks later that day
5. "Approaching weekly usage limit" banner appears while panel shows 27% / 13%
6. Banner disappears on its own

### Claude Model

None

### Is this a regression?

Yes, this worked in a previous version

### Last Working Version

_No response_

### Claude Code Version

1.46388.4

### Platform

Anthropic API

### Operating System

Windows

### Terminal/Shell

Terminal.app (macOS)

### Additional Information

### Why this looks like a metering issue, not normal usage

1. **The jump was not gradual.** Normal usage accumulates incrementally. Mine went 0% → ~60% after two short conversations.
2. **The banner contradicted the panel.** 27% / 13% is nowhere near a limit.
3. **The banner self-resolved**, suggesting a display or accounting inconsistency rather than genuine quota exhaustion.
4. This pattern resembles the widespread drain reported in #41930 (March 2026), where users saw jumps such as 21% → 100% on a single prompt, later attributed in part to two prompt-caching bugs inflating token costs 10–20×.
### Environment

- Plan: Max (5x)
- Date: September 8, 2026
- Models: Fable 5.1, plus Claude Code sessions
- Interfaces: claude.ai web, Claude Code
- Note: weekly limit was showing as "temporarily boosted (50% higher through September 13)" at the time

### What Would Help

Image

Image

Contributor guide

No contributing guide indexed for this repository

Research direction

No repository files, tests, or entry points are identified. Start by investigating the usage-metering and warning-banner paths using the reported September 8 sequence, displayed percentages, and banner state. Done means usage tracks actual consumption and the warning agrees with the usage panel without resolving incorrectly.

Written by the indexing model from the issue text.

Assessment

Domain
backend
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Active
Clarity
Needs clarification
Newbie friendliness
30/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.