z-image-turbo-sd

Curated single-volume bundle of Z-Image-Turbo for stable-diffusion.cpp serving on enclave.host (Tinfoil Modelwrap volume; the enclave attestation commits to these exact bytes).

Generic component filenames (diffusion.gguf / llm.gguf / vae.safetensors): the node's ENCLAVE_SD_*_FILE envs are node-global and must resolve inside every SD volume on a node, so all curated volumes share these relative names to coexist (matches EnclaveHost/qwen-image-2512-sd). The bytes are unchanged from the original model-specific-named revision 07cb261e โ€” this is a pure rename.

file sha256 what / source
diffusion.gguf df1c5baa86d1398c979495a6072dbcee79444fdb884a2445582ba0769c44e9a1 Z-Image-Turbo 6B Q8_0 โ€” leejet/Z-Image-Turbo-GGUF
llm.gguf 391c1e410fd9f4cf2de2b510273b56a84c19ce18f4fa3bfb3774031dac4ef068 Qwen3-4B-Instruct-2507 Q8_0 text encoder โ€” unsloth/Qwen3-4B-Instruct-2507-GGUF
vae.safetensors afc8e28272cd15db3919bacdb6918ce9c1ed22e96cb12c4d5ed0fba823529e38 FLUX.1 AE โ€” Comfy-Org/z_image_turbo

All components Apache-2.0 (Z-Image: Tongyi-MAI; Qwen3-4B: Alibaba; FLUX.1 AE: Black Forest Labs).

Serving env (enclave tinfoil-config, shared across all SD volumes on the node): ENCLAVE_SD_DIFFUSION_FILE=diffusion.gguf, ENCLAVE_SD_LLM_FILE=llm.gguf, ENCLAVE_SD_VAE_FILE=vae.safetensors, ENCLAVE_SD_VAE_TILING=64:0.25. ~13 GB peak VRAM at 1024 px, cfg 1.0, 4-8 steps, sizes in multiples of 64.

Downloads last month
131
GGUF
Model size
6B params
Architecture
Hardware compatibility
Log In to add your hardware

We're not able to determine the quantization variants.

Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for EnclaveHost/z-image-turbo-sd

Quantized
(63)
this model