# qwenImage21INT4W4A8_int4W4A8.safetensors

> Qwen Image 2.1 INT4 (W4A8) · Low-VRAM 4-bit Qwen Image 2.1

qwenImage21INT4W4A8_int4W4A8.safetensors is the Civitai download name of qwen_image_2.1W4A8.safetensors, a community INT4 (W4A8) quantization of the base Qwen Image 2.1 model by tsolful. Put it in ComfyUI/models/diffusion_models and load it with Load Diffusion Model.

W4A8 means 4-bit weights and 8-bit activations. It is the smallest Qwen Image 2.1 build that keeps the base (non-Turbo) model.

## Facts

- **Same file as:** qwen_image_2.1W4A8.safetensors
- **Type:** Diffusion model (image, community quant)
- **Format:** Safetensors · INT4 weights, INT8 activations
- **Size:** 4.20 GB (4,203,379,136 bytes)
- **SHA256:** e3312ad345f8d8a2acdb8aec25dbe4f5890a9753d30a3cd941eea80b9c85b9be
- **ComfyUI folder:** ComfyUI/models/diffusion_models/
- **Loader node:** Load Diffusion Model
- **License:** Qwen Research License
- **Released:** Sep 20, 2026 · Civitai upload

## What is qwenImage21INT4W4A8_int4W4A8.safetensors?

tsolful’s Civitai listing packs several quantizations of Qwen Image 2.1 and its Turbo variant: INT4 (W4A8), INT6 (W6A8) and INT8. This file is the base model at INT4 (W4A8), published on September 20, 2026, before the Turbo builds were added.

At 4.20 GB it is about 30% of the bf16 model, the lightest way to run the base Qwen Image 2.1. Expect some loss of fine detail and text accuracy compared with int8 or bf16.

## Download and install

- [Civitai · Qwen Image 2.1/Turbo INT4(W4A8)/INT6(W6A8)/INT8](https://civitai.com/models/2951557?modelVersionId=3342865) · saves as `qwenImage21INT4W4A8_int4W4A8.safetensors` · The “Base INT4 (W4A8)” version. Sign-in may be required.

```sh
cd ComfyUI/models/diffusion_models
# Download it from Civitai · Qwen Image 2.1/Turbo INT4(W4A8)/INT6(W6A8)/INT8, then save it here as:
# qwenImage21INT4W4A8_int4W4A8.safetensors
sha256sum qwenImage21INT4W4A8_int4W4A8.safetensors
# e3312ad345f8d8a2acdb8aec25dbe4f5890a9753d30a3cd941eea80b9c85b9be
```

Safetensors files hold tensors only and cannot run code. Still compare the SHA256: a failed download often leaves an HTML error page saved under the model’s filename.

## How to load it in ComfyUI

Load Diffusion Model reads the diffusion model and outputs a MODEL for the sampler. The text encoder and VAE load in their own nodes.

- **Also load:** A Qwen3-VL 8B text encoder (qwen3vl_8b_*) in Load CLIP and qwen_image_2.1_vae_bf16 in Load VAE.
- **Faster sampling:** Use the Turbo model, or add a Turbo/5-8 step LoRA and set CFG to 1.

## Files a Qwen Image 2.1 workflow needs

- [`qwen3vl_8b_int8_convrot.safetensors`](https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/text_encoders/qwen3vl_8b_int8_convrot.safetensors) · Text encoder (Qwen3-VL 8B) · `models/text_encoders/` · 9.35 GB
- [`qwen_image_2.1_vae_bf16.safetensors`](https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/vae/qwen_image_2.1_vae_bf16.safetensors) · VAE · `models/vae/` · 676 MB

## Workflows: Uses another version of this file

These workflows load a different precision or format of the same model. Select qwenImage21INT4W4A8_int4W4A8.safetensors in their Load Diffusion Model node to use this version.

- [Qwen Image 2.1 Outpainting ComfyUI | Seamless Expansion](https://www.runcomfy.com/comfyui-workflows/qwen-image-2-1-outpainting-in-comfyui)
- [Qwen Image 2.1 VNCCS PoseStudio | 3D Pose Control](https://www.runcomfy.com/comfyui-workflows/qwen-image-2-1-vnccs-pose-control)
- [Qwen Image 2.1 Inpainting | No-Offset Edits](https://www.runcomfy.com/comfyui-workflows/qwen-image-2-1-inpainting-in-comfyui-no-offset-edits)
- [Qwen Image 2.1 Multi-image Editing | Up to 10 Images Inputs](https://www.runcomfy.com/comfyui-workflows/qwen-image-2-1-multi-image-editing)
- [Qwen Image 2.1 Face Swap | Natural Head Swap](https://www.runcomfy.com/comfyui-workflows/qwen-image-2-1-face-swap-in-comfyui)
- [Qwen Image 2.1 Outfit Change | Precise One-click Try-On](https://www.runcomfy.com/comfyui-workflows/qwen-image-2-1-outfit-change-in-comfyui-reference-try-on)
- [Qwen Image 2.1 Edit ComfyUI | Precise Image Editing](https://www.runcomfy.com/comfyui-workflows/qwen-image-2-1-edit-in-comfyui-reference-guided-editing)
- [Qwen Image 2.1 ComfyUI Workflow: Text-to-Image](https://www.runcomfy.com/comfyui-workflows/qwen-image-2-1-comfyui-text-to-image)

## Workflows: Uses other Qwen Image 2.1 models

These workflows load another Qwen Image 2.1 model. Select qwenImage21INT4W4A8_int4W4A8.safetensors in their Load Diffusion Model node to use it instead.

- [Qwen Image 2.1 Turbo Edit | Fast Image Editor](https://www.runcomfy.com/comfyui-workflows/qwen-image-2-1-turbo-edit-comfyui-8-step-int8-editing)
- [Qwen Image 2.1 Turbo Text to Image | Fast Images](https://www.runcomfy.com/comfyui-workflows/qwen-image-2-1-turbo-in-comfyui-text-to-image)

## Same file, other names, and alternatives

- `qwenImage21INT4W4A8_int4W4A8.safetensors` · This page · base INT4 (W4A8) · 4.20 GB · Lowest VRAM, base model
- `qwen_image_2.1_W6A8_g32.safetensors` · Same listing · base INT6 (W6A8) · 5.72 GB · Better quality, still compact
- `qwen_image_2.1_turbo_W4A8.safetensors` · Same listing · Turbo INT4 (W4A8) · 4.44 GB · 8 steps on low VRAM
- `qwen_image_2.1_int8_convrot.safetensors` · Official · int8 convrot · 7.26 GB · Closer to full quality

## Common errors and fixes

### Value not in list: unet_name: 'qwenImage21INT4W4A8_int4W4A8.safetensors' not in [...]

Why: The workflow asks for this exact filename, but ComfyUI cannot find it in models/diffusion_models. Usually the file is missing, sits in a different folder, or was saved under another name such as qwen_image_2.1W4A8.safetensors.

Fix: Move the file into ComfyUI/models/diffusion_models/, rename it to match the workflow (or pick it again in the dropdown), then refresh the node list or restart ComfyUI.

### SafetensorError: Error while deserializing header

Why: The download is incomplete, or the browser saved an HTML sign-in or error page under the model’s filename.

Fix: Check that the file is exactly 4,203,379,136 bytes and that its SHA256 matches the value above. Download it again from Civitai · Qwen Image 2.1/Turbo INT4(W4A8)/INT6(W6A8)/INT8 if not.

## FAQ

### Is this the base model or Turbo?

The base model. The listing’s Turbo INT4 build is a separate file, `qwen_image_2.1_turbo_W4A8.safetensors`.

### Where do I put qwenImage21INT4W4A8_int4W4A8.safetensors in ComfyUI?

In `ComfyUI/models/diffusion_models/`. Then refresh the node list or restart ComfyUI so Load Diffusion Model lists the file.

### Which ComfyUI node loads qwenImage21INT4W4A8_int4W4A8.safetensors?

[Load Diffusion Model](/comfyui-nodes/ComfyUI/unet-loader). It is built into ComfyUI, so no custom nodes are needed.

### Can I use qwenImage21INT4W4A8_int4W4A8.safetensors commercially?

Not by default. Qwen Image 2.1 is released under the Qwen Research License, which allows research and evaluation only. Commercial use needs a separate license from Qwen (model-business@notice.qwencloud.com). Community quantizations inherit the same terms.

## Sources

- [Civitai model 2951557](https://civitai.com/models/2951557)
- [Comfy-Org/Qwen-Image-2.1](https://huggingface.co/Comfy-Org/Qwen-Image-2.1)
- [Qwen/Qwen-Image-2.1 (original)](https://huggingface.co/Qwen/Qwen-Image-2.1)

Canonical page: https://www.runcomfy.com/comfyui-workflows/model-files/qwenImage21INT4W4A8_int4W4A8.safetensors
