Z-Image is a 6-billion-parameter open image model from Alibaba's Tongyi Lab. On Kenerate, Z Image Base costs 3 credits and remixes an uploaded image with a strength slider.
Quick answer
Updated
Z-Image is a 6-billion-parameter open image model family from Alibaba's Tongyi Lab, built on a single-stream diffusion transformer. Z-Image-Turbo arrived in November 2025 and Z-Image in January 2026, with open weights. On Kenerate, Z Image Base costs 3 credits, renders up to 2048 px, and remixes one uploaded image with a strength slider — in our test, higher strength changed more of the image.
About the model
Facts from the maker's own announcement and repository, linked below.
| Maker | Alibaba's Tongyi Lab (Tongyi-MAI) |
|---|---|
| Architecture | 6B-parameter single-stream diffusion transformer |
| Released | Z-Image-Turbo November 2025; Z-Image January 2026 |
| Weights | Open (Apache 2.0) |
| Maker's highlights | Photorealism, English and Chinese text rendering, instruction following, fits 16 GB consumer GPUs |
| On Kenerate | Z Image Base — 3 credits, up to 2048 px, one reference with a strength slider |
Gallery
Real outputs made on Kenerate. Open any image to copy its prompt and run it yourself.
Same prompt
The same pencil sketch and prompt at three Reference strength settings. Higher means more change.
Finished fantasy oil painting of a warrior woman holding a spear with a flowing crimson cape, dramatic sunset sky, rich colors
Models
Every version in the generator, and what each is good at.
| Model | Reference images | Max size | Good for | Open |
|---|---|---|---|---|
| Z Image BaseAlibaba Tongyi Lab | 1 (with strength) | Up to 2048 px | Cheap photoreal drafts; image remix | Open |
From our tests
Z Image Base is an open model; Kenerate adds no filter beyond blocking prompts involving minors. For adult image to image, see NSFW image to image (18+).
How to
Try Z-Image on your own prompt — 25 free credits on sign-up.
In depth
Background, how to prompt it, and when to pick something else.
Most large image models use two streams for text and image. Z-Image uses a single-stream diffusion transformer at 6 billion parameters, which keeps it small enough for 16 GB consumer GPUs while producing photoreal results [1].
Pick it for cheap photoreal drafts and for image-to-image where you want control over how much changes. For polished finals switch to Seedream 5.0 Pro; for text-heavy designs, GPT Image 2.
We only publish real, verified reviews — never invented quotes. Send a line about your result (and whether we may quote you), and it will appear here.
FAQ
5 questions
An open 6-billion-parameter image model family from Alibaba's Tongyi Lab. Z-Image-Turbo was released in November 2025 and Z-Image in January 2026.
Z Image Base is 3 credits an image — about 3.5¢ on the $15 pack.
Yes, Tongyi Lab released the weights under Apache 2.0.
It appears under Advanced settings when you upload an image (10–100 %, default 75 %). In our test, 35 % kept the sketch, 60 % inked it, and 85 % turned it into a full painting.
Short text, yes — our 'OPEN LATE' neon sign came out correctly. For long text, GPT Image 2 is stronger.
Explore
Generators, design tools, photo edits, models and comparisons — all in one studio, on one credit balance.
Written by the Kenerate AI team
We run these models in the Kenerate generator. Model facts come from the makers' own pages; specs and limits were checked against our code; the examples are our own outputs. Last updated 27 September 2026. Corrections: hello@kenerateai.com.
Ready when you are
25 free credits on sign-up. No watermark.
