wbruna/Z-Image-Turbo-sdcpp-GGUF

Quantized GGUF versions of the Z-Image Turbo model by Tongyi-MAI, for use with stable-diffusion.cpp.

These are legacy, experimental quants. See https://huggingface.co/leejet/Z-Image-Turbo-GGUF for the official stable-diffusion.cpp quants.

For k-quants to use with ComfyUI, see https://huggingface.co/jayn7/Z-Image-Turbo-GGUF .

You can recreate these locally from the original files with something like:

./sd --mode convert --type q5_1 --model z_image_turbo_bf16.safetensors --output z_image_turbo-Q5_1.gguf

Model Information

See the original model card at Z-Image Turbo.

Usage

You need at least release master-385-34a6fd4.

Example command:

./sd --diffusion-model z_image_turbo-Q5_1.gguf --vae ae_bf16.safetensors --llm qwen_3_4b-Q8_0.gguf --cfg-scale 1 --steps 8 -p "an apple"

For lower quality, faster generations, you can use euler + sgm_uniform with very few steps, and/or TAE:

./sd --diffusion-model z_image_turbo-Q4_0.gguf --tae taef1.safetensors --llm qwen_3_4b-Q8_0.gguf --cfg-scale 1 --steps 5 --sampling-method euler --scheduler sgm_uniform -p "an apple"

Credits

• —Original Model: Z-Image Turbo by Tongyi-MAI
• —Safetensors models: ComfyUI
• —ae_bf16.safetensors from https://huggingface.co/Tongyi-MAI/Z-Image-Turbo/tree/main/vae
• —taef1.safetensors from https://huggingface.co/madebyollin/taef1/tree/main
• —GGUF files quantized with mainline stable-diffusion.cpp

License

Apache 2.0, same as the original: Z-Image Turbo.