Model reference · open weights
Z-Image-De is an open-weight image model from ostris. Z-Image-De-Turbo (BF16) weighs 12.3 GB; the smallest configuration that runs it is RTX 4090 24 GB.
Summary of the ostris/Z-Image-De-Turbo model card, 2026-10-01
What it is
| Released by | ostris |
|---|---|
| Released | 2025-12-04 |
| Parameters | 6.2B |
| VRAM | 12.3 GB for the weights |
What it runs on
| Card | One image | Memory |
|---|---|---|
| RTX 3060 12 GB … RTX 4060 Ti 16 GB | does not fit | |
| RTX 3090 24 GB | fits | 23.4 GB |
| RTX 4090 24 GB | fits | 23.4 GB |
| RTX 5090 32 GB | fits | 31.0 GB |
| L40S 48 GB | fits | 44.0 GB |
| A100 80 GB | fits | 78.2 GB |
| H100 80 GB | fits | 78.1 GB |
| RTX PRO 6000 Blackwell 96 GB | fits | 93.8 GB |
| DGX Spark (GB10) 128 GB unified | fits | 107 GB |
| H200 141 GB | fits | 138 GB |
| B200 180 GB | fits | 176 GB |
From the model card
This is a de-distilled version of Tongyi-MAI/Z-Image-Turbo. It was fine tuned on images generated by Tongyi-MAI/Z-Image-Turbo in order to break down the turbo distillation. It can be used to train LoRAs on top of that should remain compatability with the base model, or it can be continued to finetune well beyond what you could do with the turbo model with the ostris/zimage_turbo_training_adapter.
There is a ComfyUI version as well as a diffusers based version in the repo.
This can be trained on directly without an adapter. It can also be used for inference directly. For inference, use a low CFG (2.0 - 3.0) and 20-30 steps. It seems to work well with CFG normalization as well.
It wasnt released yet, as of the time of this writing, and I am impatient. :)
Quoted from the model card on Hugging Face. The full card is behind the Hugging Face link above.
Running it yourself
Rent a machine by the hour. ComfyUI is installed on it. Open ComfyUI through the tunnel, then Templates → Z-Image-Turbo: Text to Image, and choose this model's file in its model loader.
# on your rented machine (the ssh line is on its page in the console)
# get REPO FILE FOLDER: one file into /workspace/models/FOLDER, where ComfyUI loads it from
get() { hf download "$1" "$2" --local-dir /workspace/hf-files && mkdir -p "/workspace/models/$3" && mv "/workspace/hf-files/$2" "/workspace/models/$3/$4"; }
# the model (11.5 GB)
get ostris/Z-Image-De-Turbo z_image_de_turbo_v1_bf16.safetensors diffusion_models
# what the template also loads “Z-Image-Turbo: Text to Image” (7.8 GB)
get Comfy-Org/z_image_turbo split_files/text_encoders/qwen_3_4b.safetensors text_encoders
get Comfy-Org/z_image_turbo split_files/vae/ae.safetensors vae
start-comfyui# on your computer, in a second terminal: ComfyUI in your browser at http://localhost:8188
# HOST and PORT are your machine's, from its page in the console
ssh -L 8188:localhost:8188 dev@HOST -p PORT