ComfyUI v0.41.0 Released: Flux3-image and More

At a Glance
| Item | Value |
|---|---|
| Repository | Comfy-Org/ComfyUI |
| Version | v0.41.0 |
| Published | 2026-10-16 |
| License | GPL-3.0 |
| Source type | Primary source (the publisher itself) |
Values determined by this site’s code when the information was collected. Dates are JST.
Overview
ComfyUI v0.41.0 has been released. ComfyUI is a highly powerful, modular diffusion model GUI, API, and backend featuring a graph and node interface, used by many engineers as one of the fastest local inference engines in the world.
The most important change in this update is the addition of support for the latest image generation model, Flux3-image. This allows users to run advanced generation workflows using the latest models directly within ComfyUI.
Key Changes
Flux3-image Model Support
Support for the new model architecture, Flux3-image, has been added. This change directly impacts users who want to leverage the latest image generation technology in their local environments. When using the new model, please place the corresponding model file in the appropriate directory.
New Features for Image Processing and Latent Operations
Several new nodes have been introduced to enhance the expressive power of workflows.
First, the ImageSDRToHDR node has been added. This is a node for converting SDR (Standard Dynamic Range) images to HDR (High Dynamic Range). It is a useful feature for users who want to build higher dynamic range image processing or high-quality image generation workflows.
Next, the ReferenceMultiLatent node has been added. This node handles multiple latents as references, along with user experience (UX) improvements. Users performing reference-based image generation can benefit from more intuitive operations.
Additionally, a new set of nodes for acquiring the geometric structure of latents has been added. For advanced users who want to finely control and analyze shapes and structures in the latent space, this expands the workflow options.
Expanded VAE Support
Support for MiniMax-H3 LightVAE has been added. This change affects users utilizing MiniMax-related models. Utilizing a lightweight VAE enables more efficient inference and generation for specific models.
Continued Availability of LoRA Stack Nodes
During the development process, a change to temporarily disable the LoRA stack node until the frontend supports DynamicGroup was considered. However, this change was ultimately reverted. Therefore, the LoRA stack node remains available in the current version. Users can continue to manage LoRA stacking as before.
Workflow Template Updates
Provided workflow templates have been updated to v0.11.84. Users utilizing existing templates can update to the latest configuration to use workflows reflecting the newest features and optimizations.
Specifications
With this update, GGUF conversion and inference for the latest image generation model Flux3-image are now possible. In addition, MiniMax-H3 LightVAE is newly supported.
How to Get It
If you have installed it manually, please update the repository to the latest state. To install manager dependencies, run the following command:
pip install -r manager_requirements.txt
To enable the manager when running ComfyUI, start it with the following flag:
python main.py --enable-manager
Related Articles
- ComfyUI v0.44.0 Released with Expanded Partner Nodes
- ComfyUI v0.35.0 Released with Comfy Compiler and Memory Fixes
- ComfyUI v0.45.0 Released with Qwen 3.x Optimizations
- ComfyUI v0.43.0 Released: MOSS-Image and Fibo 2.0 Support
What to Read Next
- Follow this tool → ComfyUI overview and release history (163 releases tracked)
- Other image, video and audio → Diffusers / InvokeAI / whisper.cpp

