[BUG] Windows: GPU process crashes (exitCode 101457950) and app will not relaunch until Windows "Repair"

Status Closed — duplicate
Reported on v2.1.218
Maintainer reply None cached
Activity 1 comment · opened Aug 22, 2026 · closed Aug 25, 2026

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?

The Claude desktop app on Windows 11 crashes during long-running sessions. The main log records a GPU
process crash with an identical exit code every time.

The bigger problem is what happens after the crash: the app frequently will not reopen. Launching it
does nothing. The only way I have found to recover is going into Windows advanced options for the app
and selecting "Repair". Windows shows this message:

"You'll need to go to advanced options for Claude and select Repair. If you're still having issues
with it, reinstall this app or contact your administrator."

This has happened several times now, always in the middle of a task, and it kills background work that
was in flight.

What Should Happen?

The GPU process should not crash during normal use.

If it does crash, the app should recover on its own (Electron can respawn the GPU process, or fall back
to software rendering) instead of leaving the installation in a state where the app will not launch at
all until a Windows "Repair" is run. Needing an OS-level Repair to recover from an application crash is
not a reasonable recovery path for an end user.

Error Messages/Logs

From %LOCALAPPDATA%\Claude\Logs\main.log. The same block appears at every crash:

[info] GPU process gone: {
  type: 'GPU',
  reason: 'crashed',
  exitCode: 101457950,
  serviceName: 'GPU'
}

The exit code is identical every time: 101457950 decimal, 0x60C201E hex.

FREQUENCY
In a 27 hour window (2026-08-21 08:31 through 2026-08-22 11:30) the log records 6 app starts and
3 GPU process crashes:

2026-08-21 08:31:56  app start
2026-08-21 10:23:04  GPU process crashed
2026-08-21 10:28:58  app start
2026-08-21 10:29:25  GPU process crashed (only 27 seconds after start)
2026-08-21 10:37:01  app start
2026-08-21 12:01:30  app start
2026-08-22 10:00:07  app start
2026-08-22 11:24:43  GPU process crashed
2026-08-22 11:27:44  app start

One crash happened 27 seconds after the app started, so this does not look like a slow resource leak.

MEMORY RULED OUT
Process memory telemetry from the same log at the time of the crash:

[process-memory] tree_rss_sum=1848MB tree_footprint_sum=888MB
top=[electron_renderer:451MB electron_main:340MB electron_gpu:245MB]
sys_free=15937MB/32473MB

Roughly 1.8 GB in use against 15.9 GB free of 32 GB total, so this is not an out of memory condition.

WHAT I ALREADY CHECKED AND RULED OUT
A warning line, "[PreviewContext] Blocked subresource to private-resolving host", appears one to two
seconds before each crash. It is not a reliable precursor: it appears 53 times in the log against only
3 crashes, so it is most likely unrelated.

No crash dump for Claude was written to %LOCALAPPDATA%\CrashDumps. That folder contains dumps for other
applications but none for Claude.

ENVIRONMENT
OS: Windows 11 Home, version 10.0.26200.9168
Claude desktop app: 1.34493.0 and 1.34493.1 (crash occurred on both versions)
Claude Code CLI: 2.1.218
RAM: 32 GB total, about 15.9 GB free at time of crash

I can provide the full main.log if it is useful.

Steps to Reproduce

I do not have a deterministic reproduction. What is consistent about the sessions where it happens:

  1. Open the Claude desktop app on Windows 11.
  2. Start a long-running session that spawns several background tasks (subagents) running in parallel.
  3. Let it run for a while, typically while the session is doing web research across many pages.
  4. At some point the GPU process crashes and the app closes.
  5. Try to reopen the app. It often will not launch.
  6. Recover by going to Windows advanced options for the app and selecting "Repair".

The crash is not tied to a single app version: it happened on both 1.34493.0 and 1.34493.1.

Claude Model

Opus

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.218 (Claude Code)

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

PowerShell

Additional Information

Note on the environment field: I selected PowerShell because that is the shell Claude Code uses on
Windows, but I am not running Claude Code in a standalone terminal window. I am running it inside the
Claude desktop app for Windows (version 1.34493.0 / 1.34493.1).

This distinction matters for this bug: the process that crashes is the Electron GPU process, which only
exists in the desktop app. The crash is not happening in a terminal emulator.

View original on GitHub ↗

This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗