After Last Update ComfyUI is not launching anymore

[niksalyan]

What happened?

After Last Update ComfyUI is not launching anymore
Tied to reinstall SM from scratch issue persists.

Steps to reproduce

No response

Relevant logs

Version

v2.15.5

ComfyUI version v0.7.0

What Operating System are you using?

Windows

[Jascyc]

Me too.

The launch stops at last line below:
Adding extra search path checkpoints C:\AI\Data\Models\StableDiffusion
Adding extra search path diffusers C:\AI\Data\Models\Diffusers
Adding extra search path loras C:\AI\Data\Models\Lora
Adding extra search path loras C:\AI\Data\Models\LyCORIS
Adding extra search path clip C:\AI\Data\Models\TextEncoders
Adding extra search path clip_vision C:\AI\Data\Models\ClipVision
Adding extra search path embeddings C:\AI\Data\Models\Embeddings
Adding extra search path vae C:\AI\Data\Models\VAE
Adding extra search path vae_approx C:\AI\Data\Models\ApproxVAE
Adding extra search path controlnet C:\AI\Data\Models\ControlNet
Adding extra search path controlnet C:\AI\Data\Models\T2IAdapter
Adding extra search path gligen C:\AI\Data\Models\GLIGEN
Adding extra search path upscale_models C:\AI\Data\Models\ESRGAN
Adding extra search path upscale_models C:\AI\Data\Models\RealESRGAN
Adding extra search path upscale_models C:\AI\Data\Models\SwinIR
Adding extra search path hypernetworks C:\AI\Data\Models\Hypernetwork
Adding extra search path ipadapter C:\AI\Data\Models\IpAdapter
Adding extra search path ipadapter C:\AI\Data\Models\IpAdapters15
Adding extra search path ipadapter C:\AI\Data\Models\IpAdaptersXl
Adding extra search path prompt_expansion C:\AI\Data\Models\PromptExpansion
Adding extra search path ultralytics C:\AI\Data\Models\Ultralytics
Adding extra search path ultralytics_bbox C:\AI\Data\Models\Ultralytics\bbox
Adding extra search path ultralytics_segm C:\AI\Data\Models\Ultralytics\segm
Adding extra search path sams C:\AI\Data\Models\Sams
Adding extra search path diffusion_models C:\AI\Data\Models\DiffusionModels
[START] Security scan
[DONE] Security scan

ComfyUI-Manager: installing dependencies done.

** ComfyUI startup time: 2025-12-31 14:35:11.938
** Platform: Windows
** Python version: 3.12.11 (main, Jul 23 2025, 00:32:20) [MSC v.1944 64 bit (AMD64)]
** Python executable: C:\AI\Data\Packages\ComfyUI\venv\Scripts\python.exe
** ComfyUI Path: C:\AI\Data\Packages\ComfyUI
** ComfyUI Base Folder Path: C:\AI\Data\Packages\ComfyUI
** User directory: C:\AI\Data\Packages\ComfyUI\user
** ComfyUI-Manager config path: C:\AI\Data\Packages\ComfyUI\user__manager\config.ini
** Log path: C:\AI\Data\Packages\ComfyUI\user\comfyui.log

Prestartup times for custom nodes:
0.0 seconds: C:\AI\Data\Packages\ComfyUI\custom_nodes\rgthree-comfy
0.0 seconds: C:\AI\Data\Packages\ComfyUI\custom_nodes\ComfyUI-Easy-Use
5.2 seconds: C:\AI\Data\Packages\ComfyUI\custom_nodes\ComfyUI-Manager

Checkpoint files will always be loaded safely.

[NeuralFault]

go to SM packages, 3dot menu for ComfyUI > Python Packages > click the + button and paste in after the "pip install":

torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128 --force-reinstall

---

The change to cu130 PyTorch builds is causing a problem, likely an underlying compatibility conflict

---

Issue found. People are not keeping their nvidia drivers relatively up to date. Please update your drivers so that it contains support for CUDA 13 (post-August 2025).

[Jascyc]

It works now and thanks for your advice.

[niksalyan]

Yep works. Thanks !!!

[s-andrianov]

go to SM packages, 3dot menu for ComfyUI > Python Packages > click the + button and paste in after the "pip install":

Thank you very much!

[Jascyc]

Although ComfyUI can now be launched, Zimage output is just noise, is it also due to Torch compatibility conflict? Everything worked fine before the Stability Matrix update

[IdiotLol32]

go to SM packages, 3dot menu for ComfyUI > Python Packages > click the + button and paste in after the "pip install":

torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128 --force-reinstall

The change to cu130 PyTorch builds is causing a problem, likely an underlying compatibility conflict

Doesn't work for me.
It gives:
Traceback (most recent call last):
File "runpy.py", line 196, in run_module_as_main
File "runpy.py", line 86, in run_code
File "D:\StabilityMatrix\Data\Packages\ComfyUI\venv\lib\site-packages\pip_main.py", line 24, in
sys.exit(main())
File "D:\StabilityMatrix\Data\Packages\ComfyUI\venv\lib\site-packages\pip_internal\cli\main.py", line 78, in main
command = create_command(cmd_name, isolated=("--isolated" in cmd_args))
File "D:\StabilityMatrix\Data\Packages\ComfyUI\venv\lib\site-packages\pip_internal\commands_init.py", line 121, in create_command
module = importlib.import_module(module_path)
File "importlib_init.py", line 126, in import_module
File "", line 1050, in _gcd_import
File "", line 1027, in _find_and_load
File "", line 1006, in _find_and_load_unlocked
File "", line 688, in _load_unlocked
File "", line 883, in exec_module
File "", line 241, in _call_with_frames_removed
File "D:\StabilityMatrix\Data\Packages\ComfyUI\venv\lib\site-packages\pip_internal\commands\install.py", line 22, in
import pip._internal.self_outdated_check # noqa: F401
File "D:\StabilityMatrix\Data\Packages\ComfyUI\venv\lib\site-packages\pip_internal\self_outdated_check.py", line 21, in
from pip._internal.index.package_finder import PackageFinder
File "D:\StabilityMatrix\Data\Packages\ComfyUI\venv\lib\site-packages\pip_internal\index\package_finder.py", line 38, in
from pip.internal.req import InstallRequirement
File "D:\StabilityMatrix\Data\Packages\ComfyUI\venv\lib\site-packages\pip_internal\req_init.py", line 8, in
from pip._internal.cli.progress_bars import BarType, get_install_progress_renderer
File "D:\StabilityMatrix\Data\Packages\ComfyUI\venv\lib\site-packages\pip_internal\cli\progress_bars.py", line 23, in
from pip._internal.req.req_install import InstallRequirement
File "D:\StabilityMatrix\Data\Packages\ComfyUI\venv\lib\site-packages\pip_internal\req\req_install.py", line 37, in
from pip._internal.operations.build.metadata_legacy import (
ModuleNotFoundError: No module named 'pip._internal.operations.build.metadata_legacy'
Error: StabilityMatrix.Core.Exceptions.ProcessException: Process python failed with exit-code 1.
at StabilityMatrix.Core.Processes.ProcessRunner.WaitForExitConditionAsync(Process process, Int32 expectedExitCode, CancellationToken cancelToken)
at StabilityMatrix.Core.Models.PackageModification.PipStep.ExecuteAsync(IProgress1 progress) at StabilityMatrix.Core.Models.PackageModification.PipStep.ExecuteAsync(IProgress1 progress)
at StabilityMatrix.Core.Models.PackageModification.PackageModificationRunner.ExecuteSteps(IEnumerable`1 steps)

[NeuralFault]

Figured out the crux of the issue, people are just not keeping their Nvidia drivers relatively up to date.
cu130 is the current default build of stable PyTorch for CUDA/Nvidia GPUs.
SM 2.15.5 updated the default PyTorch index to cu130 for installations of ComfyUI to keep in-line with that and with ComfyUI's recommendations.

The common factor found between users coming upon this issue has been found that their Nvidia GPU drivers are too old and need to be updated due to their current installed drivers not having the CUDA 13.0 runtime needed to operate the cu130 build of PyTorch.

[ValchanOficial]

go to SM packages, 3dot menu for ComfyUI > Python Packages > click the + button and paste in after the "pip install":

torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128 --force-reinstall

The change to cu130 PyTorch builds is causing a problem, likely an underlying compatibility conflict

Thank you!!

[NeuralFault]

go to SM packages, 3dot menu for ComfyUI > Python Packages > click the + button and paste in after the "pip install":

torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128 --force-reinstall

The change to cu130 PyTorch builds is causing a problem, likely an underlying compatibility conflict

Thank you!!

Or just update your Nvidia drivers, downgrading to cu128 is a temporary workaround I found until the core issue was figured out. Relying on that just delays the inevitable.

Cause it will come a time when CUDA 13 starts being the primarily targeted dependency, and ComfyUI package force updates it cu130 build of pytorch whenever you update the package anyways.

[vldevel]

Please update your drivers so that it contains support for CUDA 13 (post-August 2025).

Sadly, even with v595, there are bugs that make anything above v575 drivers a show-stopper for me (vertices corruptions when rendering with OpenGL, for example, among many other serious bugs), so updating to v580 or newer is a no-no for now, and PyTorch v12.8 is therefore the last usable version...