Preflight Checklist
What's Wrong?
The Claude desktop app (Windows) shows a usage banner:
You've used 87% of your Opus limit. Resets Mon, Sep 28, 11:00 AM
There is no Opus-specific limit on my plan (Max 20x). Settings → Usage shows only three meters, and none of them is Opus:
- Current session: 3%
- This week (all models): 63%
- Fable this week ("Separate weekly limit for Fable"): 88%
The banner's percentage and reset time match the Fable meter, so the banner appears to read the Fable bucket and label it "Opus".
This misleads users into thinking Opus has its own cap, or that Opus usage is being charged to Fable (see #96035). I nearly commented on that issue before checking the CLI. It can also push people to stop using Opus when only the Fable cap is close.
What Should Happen?
The banner names the meter it actually reads, e.g. "You've used 87% of your Fable weekly limit", matching Settings → Usage and the CLI /usage output.
Error Messages/Logs
CLI `claude -p "/usage"` at the same time (no Opus limit listed):
Current session: 4% used · resets Sep 26, 12:10pm (Europe/Vienna)
Current week (all models): 64% used · resets Sep 28, 11am (Europe/Vienna)
Current week (Fable): 88% used · resets Sep 28, 11am (Europe/Vienna)
Steps to Reproduce
- On a Max plan, use Fable 5.1 heavily enough to approach the Fable weekly cap (here: 88%).
- Keep working in the Claude desktop app (Code tab).
- The banner appears: "You've used 87% of your Opus limit".
- Open Settings → Usage: there is no Opus meter; the matching number is "Fable this week: 88%".
- Run
claude -p "/usage" in the CLI: same numbers, no Opus limit.
Claude Model
Not sure / Multiple models
Is this a regression?
I don't know
Last Working Version
No response
Claude Code Version
CLI 2.1.283 (Claude Code); Claude for Windows desktop app 2.9939.2 (d3e504), built Sep 24, 2026
Platform
Anthropic API
Operating System
Windows
Terminal/Shell
Other
Additional Information
- The underlying numbers look correct: local transcripts show heavy Fable 5.1 use this week (≈11.9M output tokens vs ≈9.6M on Opus 5.5), so 88% of the Fable cap is plausible. Only the banner label is wrong.
- The bug is in the desktop app UI, not the CLI; filing here since the desktop Code tab ships Claude Code.
- Screenshots of the banner and Settings → Usage attached below.

Preflight Checklist
What's Wrong?
The Claude desktop app (Windows) shows a usage banner:
There is no Opus-specific limit on my plan (Max 20x). Settings → Usage shows only three meters, and none of them is Opus:
The banner's percentage and reset time match the Fable meter, so the banner appears to read the Fable bucket and label it "Opus".
This misleads users into thinking Opus has its own cap, or that Opus usage is being charged to Fable (see #96035). I nearly commented on that issue before checking the CLI. It can also push people to stop using Opus when only the Fable cap is close.
What Should Happen?
The banner names the meter it actually reads, e.g. "You've used 87% of your Fable weekly limit", matching Settings → Usage and the CLI
/usageoutput.Error Messages/Logs
Steps to Reproduce
claude -p "/usage"in the CLI: same numbers, no Opus limit.Claude Model
Not sure / Multiple models
Is this a regression?
I don't know
Last Working Version
No response
Claude Code Version
CLI 2.1.283 (Claude Code); Claude for Windows desktop app 2.9939.2 (d3e504), built Sep 24, 2026
Platform
Anthropic API
Operating System
Windows
Terminal/Shell
Other
Additional Information