README.md
8.3 KB · 127 lines · markdown Raw
1 ---
2 license: other
3 license_name: qwen-research
4 base_model:
5 - Qwen/Qwen-Image-2.1
6 base_model_relation: quantized
7 pipeline_tag: text-to-image
8 library_name: gguf
9 tags:
10 - gguf
11 - qwen
12 - image-generation
13 - comfyui
14 - comfyui-gguf
15 - mlx
16 - fp8
17 - nvfp4
18 ---
19
20 [![🤗 Model Requests](https://img.shields.io/badge/%F0%9F%A4%97_Model_Requests-Submit_Here-f59e0b?style=for-the-badge&labelColor=1f2937)](https://abenzerps.com/request-a-model/)
21
22 # Qwen-Image-2.1 Uncensored Quantized
23
24 GGUF, Safetensors, and MLX quantized formats of [Qwen/Qwen-Image-2.1](https://huggingface.co/Qwen/Qwen-Image-2.1) for local image generation using the original upstream base weights.
25
26 ## Benchmark
27
28 ![Qwen-Image-2.1 benchmark](assets/Qwen-Image-2.1-Benchmark.png)
29
30 ## Uncensored Quantized Files
31
32 | **Quantization** | **File** | **Size** |
33 | --- | --- | ---: |
34 | BF16 | [**qwen-image-2.1-UC-BF16.gguf**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/main/qwen-image-2.1-UC-BF16.gguf) | 14.23 GB |
35 | FP8 | [**qwen-image-2.1-UC-fp8.safetensors**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/main/qwen-image-2.1-UC-fp8.safetensors) | 6.63 GB |
36 | INT8 ConvRot | [**qwen-image-2.1-UC-int8_convrot.safetensors**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/main/qwen-image-2.1-UC-int8_convrot.safetensors) | 6.76 GB |
37 | NVFP4 | [**qwen-image-2.1-UC-NVFP4.safetensors**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/main/qwen-image-2.1-UC-NVFP4.safetensors) | 4.20 GB |
38 | MLX 4-bit | [**qwen-image-2.1-UC-MLX-4bit.safetensors**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/main/qwen-image-2.1-UC-MLX-4bit.safetensors) | 4.00 GB |
39 | MLX 6-bit | [**qwen-image-2.1-UC-MLX-6bit.safetensors**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/main/qwen-image-2.1-UC-MLX-6bit.safetensors) | 5.78 GB |
40 | MLX 8-bit | [**qwen-image-2.1-UC-MLX-8bit.safetensors**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/main/qwen-image-2.1-UC-MLX-8bit.safetensors) | 7.56 GB |
41 | Q8_0 | [**qwen-image-2.1-UC-Q8_0.gguf**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/main/qwen-image-2.1-UC-Q8_0.gguf) | 7.59 GB |
42 | Q6_K | [**qwen-image-2.1-UC-Q6_K.gguf**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/main/qwen-image-2.1-UC-Q6_K.gguf) | 5.88 GB |
43 | Q5_K_M | [**qwen-image-2.1-UC-Q5_K_M.gguf**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/main/qwen-image-2.1-UC-Q5_K_M.gguf) | 5.22 GB |
44 | Q4_K_M | [**qwen-image-2.1-UC-Q4_K_M.gguf**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/main/qwen-image-2.1-UC-Q4_K_M.gguf) | 4.60 GB |
45 | Q4_0 | [**qwen-image-2.1-UC-Q4_0.gguf**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/main/qwen-image-2.1-UC-Q4_0.gguf) | 4.15 GB |
46
47 **Q4_K_M** is recommended for the best balance of size and quality for GGUF. **INT8 ConvRot** and **NVFP4** offer high performance on supported GPUs, while **MLX** formats are optimized for Apple Silicon.
48
49 ## GGUF files
50
51 | **Quantization** | **File** | **Size** |
52 | --- | --- | ---: |
53 | Q8_0 | [**qwen-image-2.1-Q8_0.gguf**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/base/qwen-image-2.1-Q8_0.gguf) | 7.59 GB |
54 | Q6_K | [**qwen-image-2.1-Q6_K.gguf**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/base/qwen-image-2.1-Q6_K.gguf) | 5.88 GB |
55 | Q5_K_M | [**qwen-image-2.1-Q5_K_M.gguf**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/base/qwen-image-2.1-Q5_K_M.gguf) | 5.22 GB |
56 | Q4_K_M | [**qwen-image-2.1-Q4_K_M.gguf**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/base/qwen-image-2.1-Q4_K_M.gguf) | 4.60 GB |
57 | Q4_0 | [**qwen-image-2.1-Q4_0.gguf**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/base/qwen-image-2.1-Q4_0.gguf) | 4.05 GB |
58
59
60 ## Text Encoders & VAE
61
62 Companion model files packaged for ComfyUI:
63
64 | **Type** | **File** | **Precision** | **Size** |
65 | --- | --- | --- | ---: |
66 | Text Encoder | [**text_encoders/qwen3vl_8b_bf16.safetensors**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/main/text_encoders/qwen3vl_8b_bf16.safetensors) | BF16 | 17.53 GB |
67 | Text Encoder | [**text_encoders/qwen3vl_8b_int8_convrot.safetensors**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/main/text_encoders/qwen3vl_8b_int8_convrot.safetensors) | Int8 | 9.35 GB |
68 | VAE | [**vae/qwen_image_2.1_vae_bf16.safetensors**](https://huggingface.co/abenzerps/Qwen-Image-2.1-Uncensored-Quantized/blob/main/vae/qwen_image_2.1_vae_bf16.safetensors) | BF16 | 676 MB |
69
70 ## Usage
71
72 Use the model with [ComfyUI](https://github.com/comfyanonymous/ComfyUI) and [ComfyUI-GGUF](https://github.com/leejet/ComfyUI-GGUF).
73
74 All required companion files (GGUF transformer, text encoder, and VAE) are hosted directly in this repository.
75
76 ### 1. Download & File Placement
77
78 Download the files and place them in their respective ComfyUI directories:
79
80 ```text
81 ComfyUI/
82 └── models/
83 ├── diffusion_models/
84 │ └── qwen-image-2.1-UC-Q4_K_M.gguf # Choose one GGUF quantization (Q4_K_M recommended) or .safetensors
85 ├── text_encoders/
86 │ └── qwen3vl_8b_bf16.safetensors # Or qwen3vl_8b_int8_convrot.safetensors (recommended for lower memory)
87 └── vae/
88 └── qwen_image_2.1_vae_bf16.safetensors
89 ```
90
91 ### 2. ComfyUI Setup
92
93 1. **Install ComfyUI-GGUF**: Use the maintained fork with native Qwen-Image 2.1 support by cloning [leejet/ComfyUI-GGUF](https://github.com/leejet/ComfyUI-GGUF) into your custom nodes:
94 ```bash
95 cd ComfyUI/custom_nodes
96 git clone https://github.com/leejet/ComfyUI-GGUF
97 ```
98 *(Note: If you have the older `city96/ComfyUI-GGUF` installed and encounter an `Unknown model architecture!` error, update to the `leejet` fork above or add `ModelQwenImage` to `tools/convert.py`).*
99 2. **Node Configuration**:
100 - **Diffusion Model**: Add the **`Unet Loader (GGUF)`** node (for `.gguf` files) or standard **`UNETLoader`** (for `.safetensors` files) and select your downloaded file.
101 - **Text Encoder**: Add the standard **`CLIPLoader`** node, select `qwen3vl_8b_bf16.safetensors` (or `int8`), and set **`type`** to **`qwen_image`**.
102 - **VAE**: Add the standard **`VAELoader`** node and select `qwen_image_2.1_vae_bf16.safetensors`.
103 3. **Official Workflows**:
104 - You can use the official Comfy-Org workflow templates: [Text-to-Image](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/image_qwen_image_2_1_t2i.json) or [Image Edit](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/image_qwen_image_2_1_image_edit.json).
105 - In the workflow, simply replace the default `UNETLoader` node with **`Unet Loader (GGUF)`** when using GGUF files.
106
107 ### Memory & Performance Notes
108
109 - **Optimal Setup (GPU + RAM)**: Keep the **GGUF / Quantized diffusion model in GPU VRAM** (where speed is crucial during sampling) and let the **text encoder run in / offload to System RAM (CPU)**. Because text encoding only runs once per prompt, this saves 9–17 GB of VRAM with virtually zero impact on generation speed.
110 - **Recommended Configuration**:
111 - **Diffusion**: `qwen-image-2.1-UC-Q4_K_M.gguf` (~4.6 GB in VRAM)
112 - **Text Encoder**: `qwen3vl_8b_int8_convrot.safetensors` (~9.35 GB in RAM)
113 - **Low VRAM Mode**: If you experience VRAM out-of-memory errors, start ComfyUI with the `--lowvram` argument.
114
115 ## Uncensored
116
117 This release has no built-in safety checker or content filter. It generates adult, NSFW, and sensitive imagery directly without prompt refusals or blacked-out images. Output behavior depends solely on the input prompts and the environment in which the model is executed.
118
119 ## Source and build
120
121 - **Source model:** [Qwen/Qwen-Image-2.1](https://huggingface.co/Qwen/Qwen-Image-2.1)
122 - **Text encoder & VAE source:** [Comfy-Org/Qwen-Image-2.1](https://huggingface.co/Comfy-Org/Qwen-Image-2.1)
123 - **Source revision:** `b3179ad355be050328e483a9dfdd9e60cd62adfa`
124 - **Conversion:** [stable-diffusion.cpp](https://github.com/leejet/stable-diffusion.cpp) commit `1330cebae8f2ba99249df846cc0c9444fcbd4308` & [comfy-kitchen](https://github.com/Comfy-Org/comfy-kitchen)
125 - **License:** Qwen Research License
126 - **Checksums:** [SHA256SUMS](./SHA256SUMS)
127