New Release

Unlock Sub-Second Image Generation with Z-Image-Turbo

Name: Efficient Image Generation Model - Z-Image-Turbo
Author: Z-Image-Turbo

Experience the power of efficient image generation with Tongyi-MAI's Z-Image-Turbo. Generate photorealistic images at lightning speed, even on consumer hardware.

Generate Now Learn More

No data center required

Consumer hardware compatible

Photorealistic quality

Experience the Power of Efficient Image Generation

Z-Image-Turbo delivers unparalleled speed and quality, making professional-grade image generation accessible to all.

Ultra-Fast Generation

Generate high-quality, photorealistic images in just 8 NFEs (Number of Function Evaluations), achieving sub-second inference on enterprise H800 GPUs.

Consumer Hardware Compatible

Run Z-Image-Turbo comfortably on consumer devices with 16GB VRAM, eliminating the need for expensive data-center infrastructure.

Bilingual Text Rendering

Accurately render complex text in both English and Chinese directly within generated images, perfect for cross-market campaigns.

Generate Stunning Images in Three Simple Steps

Unleash your creativity with Z-Image-Turbo's intuitive workflow.

Enter Your Prompt

Describe the image you want to create in detail. Z-Image-Turbo understands complex descriptions.

Generate Image

Click 'Generate' and watch Z-Image-Turbo create your image in seconds, leveraging its efficient 8-step process.

Download and Share

Download your high-quality, photorealistic image and share it with the world. Use it for marketing, content creation, or personal projects.

Frequently Asked Questions about Z-Image-Turbo

Get answers to common questions about the efficient image generation model.

Z-Image-Turbo utilizes a Scalable Single-Stream DiT architecture (S3-DiT) and a novel Decoupled-DMD distillation algorithm to generate photorealistic images in just 8 NFEs.

Z-Image-Turbo requires less than 16GB of VRAM, making it suitable for consumer-grade GPUs. This contrasts with other models like FLUX.2 and Qwen-Image which have significantly higher VRAM demands.

Z-Image-Turbo matches or exceeds the performance of Seedream 4.0 while using significantly fewer parameters and computational resources. It also achieves competitive performance against Nano Banana Pro as evaluated on the Alibaba AI Arena.

The guidance scale should be set to 0.0 for Z-Image-Turbo models, as they are distilled specifically for this configuration. Setting it to other values will degrade performance.