Kromoss GGUF

This repository contains GGUF quantizations of Kromoss, a full fine-tune of MOSS-Image-Base at 1024 px. Kromoss uses the same architecture, text encoder and VAE as MOSS-Image-SFT, so it drops into the MOSS-Image workflows. The upstream model and conversion are based on the OpenMOSS release, which is licensed under Apache-2.0.

Example output generated with the supplied workflow

Quantizations

File sizes are shown in GiB. Standard quants use common GGML formats. Proprietary quants use custom ConvRot formats supported by the comfyui-gguf-reboot loader nodes.

Quant Format Size File
Q4_K_M Standard GGUF 3.35 GiB kromoss-Q4_K_M.gguf
Q6_K Standard GGUF 4.24 GiB kromoss-Q6_K.gguf
Q8_0 Standard GGUF 5.31 GiB kromoss-Q8_0.gguf
Q8_CR Proprietary, comfyui-gguf-reboot 5.04 GiB kromoss-Q8_CR.gguf
Q4_CR_L Proprietary, comfyui-gguf-reboot 3.56 GiB kromoss-Q4_CR_L.gguf

Standard GGUF files can be loaded with compatible GGUF tools. Q8_CR and Q4_CR_L use custom quantization layouts and require the comfyui-gguf-reboot loader nodes. They are not compatible with general-purpose GGUF loaders.

ComfyUI setup

Install comfyui-gguf-reboot. Use its UnetLoaderGGUFDynamicVRAM node to load a GGUF from models/diffusion_models/. These loader nodes give the best performance.

The workflow also needs the following upstream files:

Component File ComfyUI folder
Text encoder qwen3.5_9b_w4a8_gptq.safetensors models/text_encoders/
VAE flux2-vae_small.safetensors models/vae/

Download the Kromoss ComfyUI workflow.

Sampling settings

The supplied workflow uses the official Kromoss reference settings: 50 steps, CFG 3, euler sampler, sgm_uniform scheduler, and skip layer guidance on layer 22.

Inputs and outputs

Kromoss accepts natural-language prompts, tag sequences, and JSON layouts that specify text and object placement. It generates a still image. The upstream pipeline requires image dimensions that are multiples of 16.

The upstream repository also ships 8-step turbo LoRAs: lodestones/Kromoss.

License and attribution

The upstream Kromoss model is released under the Apache License 2.0. See the upstream model card and the OpenMOSS project for model details.

Downloads last month
82
GGUF
Model size
3B params
Architecture
moss_image
Hardware compatibility
Log In to add your hardware

4-bit

6-bit

8-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for molbal/Kromoss-GGUF

Quantized
(1)
this model