| File | qwen-image-Q4_K_M.gguf |
| Model | Qwen-Image (image model, 20B) |
| Precision | Q4_K_M |
| Size | 13.1 GB |
| VRAM to fit | about 15.9 GB |
| Folder | ComfyUI/models/unet/ |
| Loader node | Unet Loader (GGUF) from the ComfyUI-GGUF custom nodes |
| Source | city96/Qwen-Image-gguf on Hugging Face |
Download qwen-image-Q4_K_M.gguf (13.1 GB)
Where to put it
Put the file in ComfyUI/models/unet/ (portable version: ComfyUI_windows_portable/ComfyUI/models/unet/). Press R in ComfyUI or restart it, then pick the file in Unet Loader (GGUF) from the ComfyUI-GGUF custom nodes.
If a workflow says “Value not in list” for that node, the workflow expects a file with another name. Click the node and choose this one.
Which graphics cards it fits
| VRAM | This file | Best file for that card (RTX 40/50) | |
|---|---|---|---|
| 6 GB | Not practical | Short by 9.9 GB | qwen-image-Q2_K.gguf (7.1 GB) |
| 8 GB | Offload only | Short by 7.9 GB: runs with part of it in system RAM, slower | qwen-image-Q2_K.gguf (7.1 GB) |
| 10 GB | Offload only | Short by 5.9 GB: runs with part of it in system RAM, slower | qwen-image-Q2_K.gguf (7.1 GB) |
| 12 GB | Offload only | Short by 3.9 GB: runs with part of it in system RAM, slower | qwen-image-Q2_K.gguf (7.1 GB) |
| 16 GB | Runs | Fits in VRAM, 0.1 GB to spare | this file |
| 20 GB | Runs | Fits in VRAM, 4.1 GB to spare | qwen-image-Q6_K.gguf (16.8 GB) |
| 24 GB | Runs | Fits in VRAM, 8.1 GB to spare | qwen_image_fp8_e4m3fn.safetensors (20.4 GB) |
| 32 GB | Runs | Fits in VRAM, 16.1 GB to spare | qwen_image_fp8_e4m3fn.safetensors (20.4 GB) |
| 48 GB | Runs | Fits in VRAM, 32.1 GB to spare | qwen_image_bf16.safetensors (40.9 GB) |
Need = file × 1 + 2 GB for the picture and 0.8 GB for Windows and the browser — the same rule as the rest of the site.
Measured with this exact file
| GPU | Time | Size | Steps | Peak VRAM | Source |
|---|---|---|---|---|---|
| RTX 5060 Ti 16 GB | 123.3 s per image | 1024x1024 | 20 | 15.7 GB | measured by me, 2026-10-12 |
Time = sampling + VAE decode, after a warm-up run. Peak VRAM = memory in use on the whole card. How I measure.
What else you need
This file alone does not make a picture. With it you load:
| Role | File | Size | Folder |
|---|---|---|---|
| Model | qwen-image-Q4_K_M.gguf | 13.1 GB | models/unet |
| Text encoder | qwen_2.5_vl_7b_fp8_scaled.safetensors | 9.4 GB | models/text_encoders |
| VAE | qwen_image_vae.safetensors | 0.3 GB | models/vae |
22.7 GB in total. Smaller text-encoder versions and other options are on the Qwen-Image page.
Other files of Qwen-Image
| File | Precision | Size | VRAM to fit |
|---|---|---|---|
| qwen_image_bf16.safetensors | 16-bit | 40.9 GB | 43.7 GB |
| qwen-image-Q8_0.gguf | Q8_0 | 21.8 GB | 24.6 GB |
| qwen_image_fp8_e4m3fn.safetensors | FP8 | 20.4 GB | 23.2 GB |
| qwen-image-Q6_K.gguf | Q6_K | 16.8 GB | 19.6 GB |
| qwen-image-Q5_K_M.gguf | Q5_K_M | 14.9 GB | 17.7 GB |
| qwen-image-Q4_K_M.gguf | Q4_K_M | 13.1 GB | 15.9 GB |
| qwen-image-Q3_K_M.gguf | Q3_K_M | 9.7 GB | 12.5 GB |
| qwen-image-Q2_K.gguf | Q2_K | 7.1 GB | 9.9 GB |
Same model, different compression. 8-bit files (FP8, INT8, Q8_0) look almost the same as 16-bit; below Q5 the loss starts to show.
Free workflow with this file
- Qwen-Image — GGUF Q4_K_M — 123.3 s per 1024×1024 image
Drag the .json into ComfyUI. All free workflows →
Want the answer for your exact card? Check your GPU. Every file of every model: ComfyUI model files.
Size and source read from Hugging Face (2026-10-01).