CoolFace
Modelpublic

Rom3rk/Qwen3.8-27B-ComfyUI-GGUF

sourceHugging Faceapache-2.0updated 1mo agoView on Hugging Face
1likes3.6kdownloads
Model Card

Qwen3.8-27B-ComfyUI-GGUF

Qwen3.8-27B-ComfyUI is a fine-tuned Qwen3.8-27B model specialized for generating, editing, repairing, auditing, and explaining ComfyUI frontend/UI workflow JSON.

Overview

text
ComfyUI request
    -> interpret workflow intent and constraints
    -> generate, edit, repair, audit, or explain
    -> return workflow JSON or a conversational answer

Capabilities

  • —Generate ComfyUI UI workflows from natural-language requests
  • —Edit existing workflows while preserving requested components
  • —Repair invalid, incomplete, or disconnected workflows
  • —Audit workflow structure, nodes, inputs, and links
  • —Explain workflow behavior and node interactions
  • —Provide raw workflow JSON or conversational ComfyUI assistance

Quick Start

The examples below use Qwen3.8-27B-ComfyUI-UD-Q4_K_XL.gguf. Replace it with the GGUF file you downloaded if needed.

llama.cpp

Start an OpenAI-compatible local server:

bash
llama-server \
  -m ./Qwen3.8-27B-ComfyUI-UD-Q4_K_XL.gguf \
  --ctx-size 32768

The server listens on http://127.0.0.1:8080 by default.

Ollama

Create a file named Modelfile next to the downloaded GGUF:

dockerfile
FROM ./Qwen3.8-27B-ComfyUI-UD-Q4_K_XL.gguf
PARAMETER num_ctx 32768

Import and run it:

bash
ollama create qwen38-comfyui -f Modelfile
ollama run qwen38-comfyui

LM Studio

Import the downloaded GGUF while keeping the original file:

bash
lms import ./Qwen3.8-27B-ComfyUI-UD-Q4_K_XL.gguf --copy

You can then select and load the imported model from LM Studio.

Usage Examples

Workflow Generation

text
Create a ComfyUI UI workflow for an SDXL image-to-image pipeline with a ControlNet depth branch, a two-stage sampler, and a final image-save node. Return the complete loadable UI workflow JSON.

Workflow Editing

text
Modify the following ComfyUI UI workflow to add a second ControlNet branch. Preserve the existing model, sampler settings, node positions, and unrelated links.

Workflow Repair

text
Repair the invalid links and missing required inputs in this ComfyUI UI workflow. Keep all valid nodes and settings unchanged, and return the corrected workflow JSON.

Workflow Audit

text
Audit this ComfyUI UI workflow for invalid endpoints, disconnected nodes, incompatible inputs, redundant components, and settings that conflict with the requested generation pipeline.

Recommended Sampling Settings

ModeTemperatureTop-pTop-kPresence penalty
Thinking1.00.95200.0
Non-thinking0.70.80201.5

These settings follow the recommendations published for Qwen3.8. Availability and parameter names may vary by inference runtime.

Quantization

These GGUFs were quantized with llama.cpp using the official Unsloth Qwen3.8 Dynamic 3.0 per-tensor layouts and importance matrix. This repository is not an official Unsloth release.

Questions and Feedback

For questions, bug reports, workflow compatibility issues, or general feedback, please open a Discussion from this repository's Community tab.

When reporting a workflow issue, please include:

  • —The GGUF file used
  • —Inference runtime and version
  • —ComfyUI version
  • —Relevant custom nodes
  • —The original prompt
  • —The generated workflow JSON or raw model output

License and Attribution

This model is a modified derivative of Qwen/Qwen3.8-27B and is distributed under the Apache License 2.0. See the repository's LICENSE file for the full license text.

Qwen3.8 was developed by the Qwen Team. GGUF conversion and quantization use llama.cpp together with the official Qwen3.8 Dynamic 3.0 layouts and importance matrix published by Unsloth.

Citation

bibtex
@misc{qwen38,
  title  = {{Qwen3.8-Max}: A New Bar for Coding and Cowork},
  author = {{Qwen Team}},
  month  = {August},
  year   = {2026},
  url    = {https://qwen.ai/blog?id=qwen3.8}
}

References