Z-Image Turbo
A 6-billion-parameter model distilled to eight steps — the fastest image model on Slloy, and the cheapest at 10 credits.
10 credits, the cheapest here
The 100-credit signup bonus buys ten images instead of five. This is the model to experiment on.
Seconds per render
Eight-step distilled diffusion: the open build does 1024×1024 in about 2.3 seconds on an RTX 4090, so iteration stays conversational.
Bilingual lettering at 1024
Chinese and English text renders cleanly on posters and thumbnails, at a native 1024×1024 with 16:9, 9:16 and 3:4 crops.
Z-Image Turbo
| Native output | 1024 × 1024 · 1280 × 720 · 720 × 1280 · 864 × 1152 |
|---|---|
| Aspect ratios | 1:1 · 16:9 · 9:16 · 3:4 |
| Prompt limit | 800 characters on Slloy |
| Reference images | Not supported — text to image only |
| Price | 10 credits per image |
About Z-Image Turbo
Z-Image Turbo comes from Alibaba's Tongyi-MAI team and was released as open weights on 26 November 2025. It fits a single-stream S3-DiT architecture into 6 billion parameters and is distilled to eight sampling steps, which is why it renders a 1024×1024 frame in about 2.3 seconds on a consumer RTX 4090 — roughly an order of magnitude faster than a 20B-class model. Quality holds up better than the size suggests: portraits, product stills and social visuals come out clean, Chinese and English lettering renders reliably, and the Apache 2.0 licence means you can run the same weights yourself. On Slloy it is the cheapest image on the shelf at 10 credits, capped at 800 prompt characters, with four ratios and no reference images.
Use Z-Image Turbo for volume: drafts, A/B variants, thumbnails, social sets — anything you will regenerate ten times before it is right. Move to Seedream 5.0 when a frame has to be the final one, and to Qwen-Image when the image carries text or needs editing.
How to create images with Z-Image Turbo
- Sign up free — Create an account and get 100 free credits — no card required. That's about 5 images to test Z-Image Turbo with your own prompts.
- Write your prompt — Describe subject, style, lighting and mood. Z-Image Turbo accepts prompts in 20+ languages and handles text-in-image requests.
- Generate and download — Pick an aspect ratio, hit generate, and your 2K image is ready in seconds. Download it or keep it in your permanent gallery.
Prompt tips that work with Z-Image Turbo
- Be specific about light and lens: “golden-hour backlight, 85mm portrait” beats “nice photo” every time.
- Put text requests in quotes — “a poster that reads GRAND OPENING” — and keep the text short for clean rendering.
- Iterate: regenerate with small changes instead of rewriting the whole prompt, or edit the result with natural-language instructions.
What creators use Z-Image Turbo for
- E-commerce: product shots, lifestyle scenes and model imagery without a studio.
- Social content: scroll-stopping visuals with readable captions baked in.
- Marketing: ad variants, posters and banners in every aspect ratio you need.
- Concept art: fast style exploration across consistent characters and scenes.
FAQ
What is Z-Image Turbo?
A 6-billion-parameter open-weight text-to-image model from Alibaba's Tongyi-MAI team, released on 26 November 2025 under Apache 2.0 and distilled to eight sampling steps.
Why is it the cheapest model here?
Its cost per image is a fraction of a 20B-class model, and Slloy passes that through: 10 credits per image, so the 100 free credits on signup buy ten images.
Can I use a reference image?
No. Z-Image Turbo is text to image only — the upstream endpoint rejects reference images. Use Qwen-Image when you need to edit a photo you already have.
Is the quality good enough for real work?
For social posts, thumbnails, concepts and drafts, yes. For print-critical or brand-defining frames, step up to Seedream 5.0 at 20 credits.
Does it render Chinese text?
Yes — Chinese and English lettering is one of its strengths. Keep the prompt under 800 characters, which is the cap on Slloy.
Can I use the images commercially?
Paid plans include a commercial-use license for everything you generate. Free-plan outputs are limited to personal use — upgrade when a project goes live.
What if a generation fails?
Failed jobs are refunded automatically — credits return to your balance without a support ticket. You only pay for images that actually deliver.