Run ComfyUI at Max Speed! The Complete Guide to Python 3.13 & CUDA 13.0

Introduction

Happy New Year! I look forward to another great year with you all.

To kick off the year with a fresh start, today I’ll show you how to install ComfyUI in the latest environment to achieve maximum performance and speed.

ComfyUI: The Industry Standard for Image Generation

As of January 2026, ComfyUI has become the most widely used image generation tool for professional creators.

If you are a first-time user, I still recommend the Stability Matrix method I introduced previously, as it is very user-friendly.

Is Standalone ComfyUI More Stable?

While Stability Matrix is an excellent platform with a great UI, it isn't dedicated solely to ComfyUI.

When you install ComfyUI via Stability Matrix, certain restrictions are placed on the ComfyUI core and Manager to ensure compatibility with other tools like Stable Diffusion webUI.

flowchart LR

A1(User)

subgraph Stability Matrix

B1(Stable Diffusion webUI)
B2(ComfyUI)
B4(etc...)

end

A1-->B2
B2-->A1

D3(ComfyUI)
C1(User)

C1-->D3
D3-->C1

If you plan to use ComfyUI exclusively, a standalone installation offers better stability.

In this guide, I will show you how to keep your Stability Matrix environment intact while doing a clean install of ComfyUI in the latest environment.

The Two Essential Tools for ComfyUI

This guide assumes you are using a Windows 11 PC with an NVIDIA GPU.

To install ComfyUI, you need these two tools:

• Git: For installation and version control.
• Python: To execute the AI computations.

Let's install them one by one.

What is Git?

ComfyUI is developed on a platform called GitHub.

GitHub is a Microsoft-backed development platform designed to allow multiple developers to collaborate on projects while managing version control effectively. Git is the tool that lets you use GitHub’s features on your local computer, making the installation of ComfyUI a breeze.

Installing Git

First, go to the Git download page and grab the installer.

Run the downloaded file. During the installation, you will be asked many configuration questions—you can simply leave everything at the default/recommended settings and proceed.

Python: The Language Powering AI

Once Git is installed, the next step is Python.

Python is a programming language with a long history (dating back to 1991), but it has seen a massive surge in popularity recently due to its excellent compatibility with AI tasks.

ComfyUI Now Supports Python 3.13!

While Python releases new versions every year, ComfyUI officially added support for Python 3.13 in August 2025.

ComfyUI Supported Python Versions:

• Python 3.13: Latest, highest performance.
• Python 3.12: Stable.
• Python 3.10: Works with older extensions.

Speed Difference: Python 3.13 vs. 3.12

While newer versions generally offer better performance, some legacy extensions might not work on the latest release.

Since ComfyUI is currently being developed using Python 3.13, I recommend trying Python 3.13 first when installing. If you encounter any issues, you can then downgrade to Python 3.12.

Download Python 3.13

You can download Python 3.13 from the link below:

Run the installer. Crucially, make sure to check both boxes at the bottom of the first screen, including "Add python.exe to PATH".

All other settings can remain at their recommended defaults.

Installing ComfyUI

Now, let's install the ComfyUI core. You can install it anywhere, but I recommend a shallow directory close to the drive root.

For this guide, we will install it directly in the C: drive. Open File Explorer and go to your C: drive.

Once there, click the right side of the address bar, type cmd, and press Enter. A black Command Prompt window will open.

Copy and paste the following command, then press Enter:

git clone https://github.com/comfyanonymous/ComfyUI.git

You have now successfully installed ComfyUI using Git!

Setting Up the Python Environment

Next, we need to prepare the environment for Python to run inside the ComfyUI folder. Run the following commands one by one:

cd ComfyUI

python -m venv venv

venv\Scripts\activate.bat

After the final command, your prompt should change to show (venv) at the beginning, indicating the virtual environment is active.

Installing Python Packages

Now we will install the packages required to run ComfyUI.

Packages with CUDA 13.0 Specification

CUDA is the framework used to run AI tasks on NVIDIA GPUs. Since CUDA is generally backward compatible, we will install the packages optimized for the new CUDA 13.0.

Enter the following command into the active terminal:

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu130

Note: If you encounter CUDA-related errors, try changing cu130 to cu128 or cu126 to use a slightly older version.

Installing Remaining Dependencies

Next, install all other required packages with this command:

pip install -r requirements.txt

Launching ComfyUI!

Let’s fire it up! Enter the following command:

python main.py --auto-launch

If the browser opens and the workflow appears, you’ve succeeded!

Optimizing Your Setup

Let's make this standalone installation even more convenient.

1. Install ComfyUI-Manager

Navigate to the ComfyUI/custom_nodes folder in File Explorer. Again, type cmd in the address bar and run:

git clone https://github.com/Comfy-Org/ComfyUI-Manager.git

2. Create a Launch Batch (.bat) File

Instead of typing commands every time, let's create a shortcut. Create a file named ComfyUI_launch.bat in your main ComfyUI folder (the same place where main.py is) and paste this code:

@echo off

call venv\Scripts\activate.bat

set ARGS= --auto-launch

echo Launching ComfyUI with arguments:
echo %ARGS:^&=%
echo.

python main.py %ARGS%

pause

• ComfyUI_launch.bat

Now, you can just double-click this .bat file to start ComfyUI.

3. Share Models with Stability Matrix

If you already have models downloaded in Stability Matrix, you don't need to download them again. You can share them.

Right-click your ComfyUI_launch.bat and select Edit (Notepad). Modify the third line to include the config argument:

set ARGS= --auto-launch --extra-model-paths-config

Now, find the extra_model_paths.yaml file inside your StabilityMatrix/Packages/ComfyUI folder. Right-click it and select "Copy as path."

Paste that path into your .bat file like this:

set ARGS= --auto-launch --extra-model-paths-config "YourPath\StabilityMatrix\Packages\ComfyUI\extra_model_paths.yaml"

Next time you launch, ComfyUI will automatically load all your Stability Matrix models!

4. Custom Output Folder

By default, images go to ComfyUI/output. To change this (e.g., to your Windows "Pictures" folder), add the following argument to your .bat file:

set ARGS= --auto-launch --output-directory "C:\Users\YourName\Pictures"

Helpful Launch Arguments

• Free up VRAM (If using another GPU for display):

--reserve-vram 0

• Use FP32 Text Encoder:

--fp32-text-enc

• Use FP32 VAE:

--fp32-vae

• Disable Async Offload (If experiencing instability):

--disable-async-offload

Full list of arguments

Summary

• ComfyUI now supports Python 3.13 and CUDA 13.0.
• The latest versions offer performance gains.
• You can easily share models with Stability Matrix to save space.

By installing a simple, standalone environment, you can minimize troubleshooting and keep things running smoothly despite ComfyUI's frequent updates.

If you have the storage space, it’s a great idea to keep one "Latest" install and one "Stable" install. Why not give this high-speed environment a try?

Thanks for reading!