[BUG] Unexpected 5-hour usage consumption after Claude for Windows update

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?

Hello Claude Support,

I would like to report an issue and request clarification regarding unexpected usage consumption in Claude for Windows.

I opened Claude for Windows, and without me doing anything or loading any previous chat, the app downloaded an update by itself: Version 1.11847.5 (9692f0). After the update, the app restarted and showed the new model “Fable 5”.

Immediately after that, I asked a very simple question with the new model selected:

“How many more tokens do you consume compared with Sonnet 4.6?”

To my surprise, the context window appeared to be full and my 5-hour usage limit went up to 23%.

I am attaching a screenshot showing the usage status after the issue occurred. Please note that the screenshot was taken after I switched the chat to Sonnet 4.6 in order to investigate the reason for the unexpected consumption, because Sonnet 4.6 uses fewer tokens. Therefore, the model shown in the screenshot is not necessarily the model that caused the initial usage increase.

The screenshot shows the following information:

Context window: 235.4k / 200.0k tokens, shown as 100%
Plan usage
5-hour limit: 23%, resets in 4 hours
Weekly usage — all models: 3%, resets in 3 days
The bottom of the screenshot shows the model/status area as Sonnet 4.6 — Medium, because I had already switched models by that point

When I asked Claude why this usage had happened, it replied that the 23% usage was not caused by that single question alone, but by accumulated context from a previous session, including a previous SOAP→REST migration task, compacted context, project code, memory files, and a full claude-api skill invocation that supposedly added tens of thousands of tokens to the context.

However, I had not used Claude on any of its platforms for more than 15 hours, so the 5-hour usage window should have been completely reset. Also, this appeared to be a new session from my perspective, since I had just opened the Windows app after the update and had not manually resumed or loaded any previous conversation.

Because of this, I believe the usage consumption does not correspond to the simple question I asked, and I would appreciate clarification on what happened. Specifically, I would like to understand why the 5-hour limit showed 23% usage immediately after asking such a small question, especially after more than 15 hours of inactivity.

Please review this case, as this level of consumption for a simple question seems incorrect or at least unclear from the user’s perspective.

Thank you.

<img width="431" height="191" alt="Image" src="https://github.com/user-attachments/assets/d3e021a6-88aa-4b1a-b14c-a202bf4f9db1" />

What Should Happen?

After more than 15 hours of inactivity, the 5-hour usage limit should have been fully reset. When opening Claude for Windows after the update and asking a simple question in what appeared to be a new session, Claude should not have loaded or inherited a previous large context automatically, and the usage should not have increased to 23% from a single simple question.

If previous context was being restored or reused, the app should have clearly shown that before sending the message, including the active conversation/context size and the expected impact on usage.

Error Messages/Logs

Steps to Reproduce

no es un reclamo por consecución de acciones

Claude Model

Other

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

Versión 1.11847.5 (9692f0)

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_