Repackaged model files for ComfyUI. Place the files in the following folders
SAVRN's Notes on z_image_turbo
Plan disk before you plan GPU on this one. Comfy Org repackages these diffusion model files for ComfyUI, with the instruction to place each file in its folder, and the package weighs 40.7 GB across 11 files. We have not published a runs_on memory line for it, so there is no cheapest-card figure to quote; the disk size is what you size against, and 7.8 million downloads a month say plenty of people are doing exactly that.
Apache 2.0 covers it: commercial use, modification and redistribution, with the notices and any NOTICE file kept and significant changes stated. Before committing, follow the derivation back to Tongyi-MAI/Z-Image-Turbo, because this listing repackages that model rather than originating it, and confirm which of the 11 files your ComfyUI workflow actually loads. Released November 26, 2025 and last updated August 17, 2026; make sure the copy you pulled matches the later date.
Model Card
By Comfy Org, published under apache-2.0, revision 08d044552790.
Repackaged model files for ComfyUI. Place the files in the following folders
Read Comfy Org's full model card
Repackaged model files for ComfyUI.
Original model repository: https://huggingface.co/Tongyi-MAI/Z-Image-Turbo
Place the files in the following folders:
ComfyUI/
├── models/
│ ├── diffusion_models/
│ │ ├── z_image_turbo_bf16.safetensors
│ │ ├── z_image_turbo_int8_convrot.safetensors
│ │ └── z_image_turbo_nvfp4.safetensors
│ ├── loras/
│ │ └── z_image_turbo_distill_patch_lora_bf16.safetensors
│ ├── text_encoders/
│ │ ├── qwen_3_4b.safetensors
│ │ ├── qwen_3_4b_fp4_mixed.safetensors
│ │ └── qwen_3_4b_fp8_mixed.safetensors
│ └── vae/
│ └── ae.safetensors
Workflows: https://comfyanonymous.github.io/ComfyUI_examples/z_image/
Identity and Version
- Repository
- Comfy-Org/z_image_turbo
- Publisher
- Comfy Org
- Task
- Not stated by the source
- Modality
- Other
- Library
- diffusion-single-file
- Parameters
- Not stated by the source
- Languages
- Not stated by the source
- Revision
- 08d04455279082882deaabc8d0d09fc914c071e1
- First published
- 2025-11-26
- Last updated
- 2026-08-17
Files and Weights
11 files, 40.7 GB in total. The weights are 8 files totalling 40.7 GB in safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| split_files/diffusion_models/z_image_turbo_bf16.safetensors | Weights | 12.3 GB | 2407613050b8 |
| split_files/diffusion_models/z_image_turbo_int8_convrot.safetensors | Weights | 6.2 GB | be517ebd47c9 |
| split_files/diffusion_models/z_image_turbo_nvfp4.safetensors | Weights | 4.5 GB | a553c889dbcb |
| split_files/loras/z_image_turbo_distill_patch_lora_bf16.safetensors | Weights | 158.8 MB | 340cc573d887 |
| split_files/text_encoders/qwen_3_4b.safetensors | Weights | 8.0 GB | 6c671498573a |
| split_files/text_encoders/qwen_3_4b_fp4_mixed.safetensors | Weights | 3.5 GB | 7ca32dcf07df |
| split_files/text_encoders/qwen_3_4b_fp8_mixed.safetensors | Weights | 5.6 GB | 72450b197581 |
| split_files/vae/ae.safetensors | Weights | 335.3 MB | afc8e28272cd |
| z_image_convert_original_to_comfy.py | Configuration | 1.6 KB | — |
| README.md | Documentation | 953 B | — |
| .gitattributes | Repository | 1.5 KB | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
- Download size
- 40.7 GB
Released by Comfy Org through its official repository on Hugging Face. Read the license.
Built From
- Derived from Tongyi-MAI/Z-Image-Turbo
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 40.7 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About z_image_turbo
Can I use z_image_turbo commercially?
Yes. z_image_turbo is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.