Apps
ChatGPT Desktop Crash Reports
ChatGPT Desktop on Windows stores Chromium Crashpad minidumps and metadata in its Crashpad reports folder whenever the app process, renderer, GPU process, or helper components terminate unexpectedly. These .dmp crash dumps and companion report files exist to help diagnose repeated failures, but older reports are no longer needed for normal app operation; Kudu removes crash reports older than one week without touching chats, prompts, account sign-in, or app settings.
Why clean ChatGPT Desktop Crash Reports?
- Repeated renderer or GPU-process crashes can leave behind many Crashpad minidumps, and the first visible symptom is the ChatGPT Desktop cache folder growing unexpectedly large in LocalAppData
- Crash loops during startup or after waking from sleep generate a new report on each failure, so users often notice disk space being consumed even though the app never stays open long enough to use
- Older crash reports do not help the app launch faster or run better, so keeping weeks of .dmp files only preserves obsolete diagnostics while the user sees no benefit beyond higher storage use
- Systems with frequent graphics-driver resets can accumulate GPU-process crash dumps rapidly, which shows up as repeated report files after black windows, flickering, or sudden app closes
- Background helper crashes may be invisible during normal use, but the reports folder still fills with Crashpad artifacts and users only discover it when storage analyzers flag ChatGPT Desktop as using far more space than expected
- Cleaning out reports older than a week keeps recent diagnostic evidence for current problems while removing stale dumps from already-resolved crashes that would otherwise continue piling up
What gets cleaned
Cache paths Kudu targets
Windows
%LocalAppData%/Packages/OpenAI.ChatGPT-Desktop_2p2nqsd0c76g0/LocalCache/Roaming/ChatGPT/Crashpad/reports |
Frequently asked
Common questions about ChatGPT Desktop Crash Reports
Related cleaners
Free & open source
Download Kudu and reclaim your disk space.
Available on Windows, macOS, and Linux. No account required, no feature gates, no telemetry without consent. All cleaning targets are open source and community-auditable.