#3434th of 34 in Open modelsFor image, video and design, image and design
z-image
6B-parameter text-to-image diffusion model family with an 8-step Turbo variant
Live demo ↗
(opens in a new tab)Website ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 12.1k
- License
- Apache-2.0
- Last commit
- Feb 2026
- Language
- Python
10,629 on 31 Mar 202612,107 on 20 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Z-Image is a family of 6B-parameter image generation models built on a single-stream diffusion transformer (S3-DiT), with PyTorch and diffusers inference code. Z-Image-Turbo is a distilled variant that runs in 8 function evaluations and fits in 16 GB VRAM, while Z-Image is the 50-step base model with CFG and negative prompts. Omni-Base and Edit variants are listed as not yet released.
Who it is for: Engineers running or fine-tuning open-weight text-to-image models
Strengths
- Turbo variant needs 8 NFEs and fits in 16 GB VRAM consumer GPUs
- Base Z-Image supports CFG and negative prompts, and is aimed at fine-tuning
- Renders English and Chinese text in images
- Community ports: stable-diffusion.cpp runs it on 4 GB VRAM, plus vllm-omni and SGLang support
Weaknesses
- Omni-Base and Edit checkpoints are marked 'To be released' in the model zoo
- Turbo has low output diversity and no CFG, per the README table
- Diffusers support requires installing from source per the README
- No release tags; inference is a script plus library, not a server or UI
What it needs
- GPU required
- Needs PyTorch, diffusers
- Models: Z-Image, Z-Image-Turbo, Hugging Face safetensors checkpoints, diffusers ZImagePipeline
Also in Open models
See all 34| Rank | Project | Score |
|---|---|---|
| 1 | llmfitRust tool that checks which LLMs fit your RAM, CPU and GPU | 87 out of 100 |
| 2 | layaSingle-pass multilingual classifier for typed choice, score and yes/no decisions | 84 out of 100 |
| 3 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 75 out of 100 |
| #4 | speechPyTorch toolkit for training and deploying ASR, TTS and speech LLM models | 70 out of 100 |
| #5 | yueLyrics-to-song model that writes an editable melody and chord score first | 65 out of 100 |