Z Image Turbo (zimageturbo.ai)

Z Image Turbo is a 6B S3-DiT text-to-image model distilled to an 8-step pipeline, hosted on zimageturbo.ai. It runs on a managed GPU cluster where most scenes finish in 2–4 seconds; a mid-range self-hosted GPU typically stays under 10 seconds. The bilingual English/Chinese checkpoint runs in the browser workspace, through the REST API, or inside ComfyUI. Official Tongyi-MAI weights are Apache-licensed; this site adds credits, stored prompts, and team folders. Keep CFG at 1.0, Shift between 3 and 6, and steps at 8–10 so the distilled sampler does not collapse.

Z Image Turbo can be used directly in a browser or run locally via ComfyUI or other flows. The same model files power both online and offline paths, and users can export the weights or deploy on compatible runtimes. The hosted workspace provides free credits, while paid plans add volume. It emphasizes bilingual support and prompt coherence across languages.

Z Image Turbo offers fast production-like performance with a model designed for real-time interaction and bulk generation. It supports English and Chinese prompts and renders multilingual text directly in images. Users can access it via a browser, REST API, or ComfyUI, and can also run it on self-hosted setups with compatible GPUs.

Z Image Turbo provides a set of plans and integration options, including a hosted workspace, ComfyUI workflow compatibility, REST endpoints, and downloadable weights. It also mentions limits such as 8-step sampling and VRAM considerations for local runs.

Use cases

• Rapid concept art and product visuals in marketing decks