# Z-Image Turbo

> Z-Image Turbo is available on Artifio.ai — a unified AI workspace with a single pay-as-you-go wallet.

- Canonical page: https://artifio.ai/models/alibaba-z-image-turbo
- Run it: https://artifio.ai/create/alibaba-z-image-turbo
- Modality: prompt to image
- Model brand: Alibaba
- Pricing: from about $0.005 per generation (exact cost shown in the workspace before you confirm; varies with selected options)
- Typical generation time: ~10s (catalogue estimate)
- Full pricing table: https://artifio.ai/pricing/models

## About Z-Image Turbo

Z-Image Turbo is the distilled speed variant of Z-Image, the 6B-parameter Single-Stream Diffusion Transformer family from Tongyi-MAI, Alibaba's Tongyi lab. Released on November 26, 2025, it was the first member of the family to ship, and it compresses sampling down to 8 function evaluations through few-step distillation with reward post-training. The result is sub-second generation on an H800 datacenter GPU and a footprint that fits inside 16GB of consumer VRAM, unusual territory for a model that trades punches with 20B-plus systems.

By December 8, 2025 it ranked first among open-source models on the Artificial Analysis text-to-image leaderboard. On Artifio it is one of the fastest image models in the catalogue, running pay-per-generation from the same wallet as everything else.

## Features

- 6B-parameter S3-DiT architecture distilled to 8 NFEs (function evaluations) per image
- Sub-second inference latency on H800 GPUs and compatibility with 16GB-VRAM consumer cards
- Photorealistic generation with strong aesthetic quality, the variant's stated specialty
- Bilingual text rendering in English and Chinese
- Apache 2.0 weights with official support merged into the Diffusers library
- Part of a four-variant family alongside Z-Image (base), Z-Image-Edit, and Z-Image-Omni-Base

## What people use it for

- Rapid iteration: cycling through prompt variations at interactive speed before committing to a slower model
- High-volume asset generation where per-image latency and cost dominate
- Photorealistic portraits, which the launch material highlights as a particular strength
- Quick bilingual signage, poster, and product mockups with embedded text

## Reported strengths

- Ranked first among open-source models on the Artificial Analysis text-to-image Elo leaderboard within two weeks of release
- The technical report describes performance comparable to leading proprietary systems while using the smallest parameter count in the leaderboard's top tier
- 8-step sampling makes it dramatically cheaper to run than 28-to-50-step base models of similar quality
- Open Apache 2.0 license, with weights small enough for ordinary gaming GPUs

## Reported limitations

- Distillation lowers output diversity; the family's own comparison table marks Turbo as low-diversity next to the base Z-Image
- No classifier-free guidance and no negative prompting support, so prompt-level control is coarser than on the base checkpoint
- Not intended for fine-tuning; Tongyi-MAI points custom-training work at the undistilled Z-Image instead

## Sources

1. Tongyi-MAI/Z-Image-Turbo model card (Hugging Face (Tongyi-MAI)): https://huggingface.co/Tongyi-MAI/Z-Image-Turbo
2. Z-Image: An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformer (arXiv): https://arxiv.org/abs/2511.22699
3. Tongyi-MAI/Z-Image repository (GitHub (Tongyi-MAI)): https://github.com/Tongyi-MAI/Z-Image

---

About Artifio.ai: 100+ AI image, video, and audio models in one workspace with a single shared wallet — no per-model subscriptions. This is the markdown rendition of https://artifio.ai/models/alibaba-z-image-turbo for AI assistants and crawlers; the HTML page is canonical.