MOSS-Image Base & SFT (INT6 W6A8 & INT4 W4A8) for ComfyUI

Repackaged model files for ComfyUI with low-bit quantized versions of both MOSS-Image-Base and MOSS-Image-SFT.

The original models are:

• OpenMOSS-Team/MOSS-Image-Base
• OpenMOSS-Team/MOSS-Image-SFT

The official ComfyUI repackage (BF16 and INT8 ConvRot diffusion models, text encoders and VAE) is at Comfy-Org/MOSS-Image.

Files

GPTQ (recommended)

• moss_image_base_W6A8_GPTQ.safetensors: the MOSS-Image-Base diffusion transformer, GPTQ-quantized to W6A8 on all blocks. Highest quality in this repo.
• moss_image_sft_W6A8_GPTQ.safetensors: the MOSS-Image-SFT diffusion transformer, GPTQ-quantized to W6A8 on all blocks. Highest quality in this repo.
• moss_image_base_W4A8_GPTQ.safetensors: the MOSS-Image-Base diffusion transformer, GPTQ-quantized with mixed W4A8 / W6A8 layers. Best balance of size and quality.
• moss_image_sft_W4A8_GPTQ.safetensors: the MOSS-Image-SFT diffusion transformer, GPTQ-quantized with mixed W4A8 / W6A8 layers. Best balance of size and quality.
• moss_image_base_W4A8_GPTQ_quant_report.json / moss_image_sft_W4A8_GPTQ_quant_report.json: quantization reports from the GPTQ conversion of the mixed files (for reference only, not needed in ComfyUI).

These use GPTQ instead of plain rounding (RTN), which gives lower error at the same file size.

• W6A8 GPTQ: same size as the plain W6A8 files (4.37 GB), with about half the output error (0.042 → 0.020).
• W4A8 GPTQ (mixed): blocks 0–23 are W4A8 (group size 16). The last 18 blocks (24–41) are kept at W6A8 (group size 32). Those extra bits make the files larger than the plain W4A8 files (3.84 GB vs 3.33 GB), but output quality is much higher. Their output error (0.023) is almost as low as W6A8 GPTQ, with files 0.53 GB smaller.

See GPTQ results.

Plain rounding (RTN)

• moss_image_base_W6A8.safetensors: the MOSS-Image-Base diffusion transformer as a single file in ComfyUI's native INT6 W6A8 format.
• moss_image_base_W4A8.safetensors: the MOSS-Image-Base diffusion transformer as a single file in ComfyUI's native INT4 W4A8 (codebook) format.
• moss_image_sft_W6A8.safetensors: the MOSS-Image-SFT diffusion transformer as a single file in ComfyUI's native INT6 W6A8 format.
• moss_image_sft_W4A8.safetensors: the MOSS-Image-SFT diffusion transformer as a single file in ComfyUI's native INT4 W4A8 (codebook) format.

Text encoder and VAE

The text encoder and VAE are not included here. Download them from Comfy-Org/MOSS-Image:

• text_encoders/qwen3.5_9b_int8_convrot.safetensors or text_encoders/qwen3.5_9b_w4a8_gptq.safetensors
• vae/flux2-vae.safetensors

Model storage paths

Place the files in the following folders:

📂 ComfyUI/
├── 📂 models/
│   ├── 📂 diffusion_models/
│   │   ├── moss_image_base_W6A8_GPTQ.safetensors
│   │   ├── moss_image_base_W4A8_GPTQ.safetensors
│   │   ├── moss_image_base_W6A8.safetensors
│   │   ├── moss_image_base_W4A8.safetensors
│   │   ├── moss_image_sft_W6A8_GPTQ.safetensors
│   │   ├── moss_image_sft_W4A8_GPTQ.safetensors
│   │   ├── moss_image_sft_W6A8.safetensors
│   │   └── moss_image_sft_W4A8.safetensors
│   ├── 📂 text_encoders/
│   │   ├── qwen3.5_9b_int8_convrot.safetensors
│   │   └── qwen3.5_9b_w4a8_gptq.safetensors
│   └── 📂 vae/
│       └── flux2-vae.safetensors

You only need the diffusion model files you plan to use.

Load the diffusion model with ComfyUI's Load Diffusion Model node, the same way as the Comfy-Org files. The tensor names match Comfy-Org's repackage, and the quantization settings are stored in ComfyUI's native _quantization_metadata header, so no extra loader is needed.

The original model recommends 50 steps, CFG 3, shift 5, and negative-branch skipping at layer 22. I've been finding good results at a minimum of 25 steps, CFG 3, and negative-branch skipping at layer 22. With the shift node at 5, the results are identical to running without the node.

Conversion details

The large layers of all 42 transformer blocks (168 layers) are stored at 6 bits (W6A8) or 4 bits (W4A8) instead of 16 bits. Small, sensitive parts such as the embedders, modulation, output layer and norms stay at full precision. Converted with my ComfyUI-INT8-Fast (fork). Requires comfy-kitchen >= 0.2.37 on ComfyUI v0.38.0 or newer.

How it compares

Image prediction difference vs BF16 (SFT model, RTN files):

• W6A8: about 19 % smaller than INT8 and very close to BF16. Images should be nearly identical.
• W4A8: about 38 % smaller than INT8, with roughly 3× the error of W6A8. Expect more visible differences in layout and fine detail.
• The difference is largest at the first, noisiest step and shrinks after that.

GPTQ results

Output error is the relative L2 difference of the model output vs BF16 (lower is better).

• W6A8 GPTQ: the lowest error in this repo (0.020), about half that of plain W6A8 at the same size.
• W4A8 GPTQ mixed: keeping the last 18 blocks at W6A8 adds about 0.5 GB over plain W4A8. In return, output error drops by about 5× (0.111 / 0.123 → 0.023). That is nearly as good as W6A8 GPTQ, with files 0.53 GB smaller.
• Plain W4A8: the smallest files, but with the highest error, especially on Base (0.123).

Check output quality for your own prompts and workflows before relying on it.

License

The original model is released under the Apache License 2.0.