Windows Error 0xc000001d

[kikilloy]

Describe the bug
When installing the cu130 wheel of llama-cpp-python according to the instructions and running a test workflow, the program crashes immediately with the following error:
OSError: [WinError -1073741795] Windows Error 0xc000001d

I have tried to troubleshoot but couldn't resolve it.

To Reproduce
Steps to reproduce the behavior:

1. Install the cu130 pre-compiled wheel of llama-cpp-python.
2. Set up and run a basic test workflow (working with gemma4-e2b-qat and its mmproj).
3. See error: OSError: [WinError -1073741795] Windows Error 0xc000001d.

Expected behavior
The library should load successfully and execute the workflow without throwing an OS-level illegal instruction or DLL crash.

logs
Diagnostics-1783163849.log

workflow

test.json

Desktop (please complete the following information):

• OS: Windows 11
• Browser: edge
• Version: comfyui v0.26.0
• Environment: PyTorch 2.9.1+cu130, CUDA 13.0

Smartphone (please complete the following information):

• N/A

Additional context

• N/A

My Hardware Info:

• CPU: i7-13650HX
• GPU: 5060-8G laptop
• NVIDIA Driver Version: 591.86

[Grain-Kitty]

I encountered this exact same error too after installing the cu130 prebuilt wheel. After spending quite a while troubleshooting, I switched to a different build of llama-cpp-python and the workflow ran without crashing.
The repository link is: https://github.com/JamePeng/llama-cpp-python/
Hope this solution works for you!