Updated: Jul 5, 2026
base modelDownload
4 variants available
fp8 SafeTensor
cococomix_int8_4step_V10.safetensors
8-bit, smaller file โข 6.69 GB
Verified: 18 days ago
SafeTensor
2220 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9
(14)
Jun 26, 2026
ZImageTurbo
v1.0 int8:
int8
v1.0 fp8:
fp8
v1.0 4step int8:
4step test
License:
Apache 2.0I mixed in a few photorealistic models.
๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ๐ฆ
V2.0~ int8 models can be loaded using Comfyui's native loader! You're running the latest version of Comfyui, right!?
for ~v1.1 int8 model
ComfyUI-INT8-Fast
https://github.com/BobJohnson24/ComfyUI-INT8-Fast
About the 4step Model
alibaba-pai/Z-Image-Fun-Lora-Distill
https://huggingface.co/alibaba-pai/Z-Image-Fun-Lora-Distill
I tried merging the 4steplora for ZIB.
I know that ZIT weights arenโt used, but I felt it might be effective when actually applied, so I gave it a try.
RTX 3060 with 12GB of VRAM, it can generate a 1024ร1024 image in under 10 seconds.
0.5Mpixels
1Mpixels
Lora might not be working right.
sample
w 4step LORA

w/o 4step LORA



