[BUG] Desktop (macOS): renderer pegs 100%+ CPU on modal dialog open from GUI
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?
- Claude Desktop for macOS — version
1.19367.0 (1a5be1), build2026-07-07T05:45:41Z - Intel Mac, macOS Tahoe 26.3.1
- Clean install — no connectors, plugins, or integrations
- No in-app feedback option present in this build's GUI (filing here for lack of a channel)
- Onset: began after a recent app auto-update
What Should Happen?
Opening a modal should be near-instant and should not saturate a CPU core.
Error Messages/Logs
- The `Claude Helper (Renderer)` process spikes to ~100% CPU the instant the modal dialog appears.
- The entire OS becomes unresponsive until the dialog is dismissed.
- Text typed into the dialog appears character-by-character with multi-second latency.
Steps to Reproduce
Just use normally, start claud.app (without any parameters, not started by terminal, just clicking on app-icon
Claude Model
Not sure / Multiple models
Is this a regression?
Yes, this worked in a previous version
Last Working Version
not sure, this behaviour now persisted throug several autoupdate cycles, but was not hitting me in 'some' past versions
Claude Code Version
1.19367.0 (1a5be1), build 2026-07-07T05:45:41Z
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
Terminal.app (macOS)
Additional Information
Working workaround (for other affected users)
Launch with both GPU flags:
/Applications/Claude.app/Contents/MacOS/Claude --disable-gpu --disable-gpu-compositing
To make it persistent, a wrapper script on PATH:
#!/bin/bash
exec /Applications/Claude.app/Contents/MacOS/Claude --disable-gpu --disable-gpu-compositing "$@"
Do not combine the Help-menu "disable GPU" toggle with a partial flag (e.g. only --disable-gpu-compositing) — that contradictory state triggers #76067
This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗