Ming-Image 0.1 Design

inclusionAI's two 6B models: one generates text-heavy designs and UI, one splits flat designs into editable layers. MIT license.

In short. Ming-Image 0.1 Design is a pair of image models for designers. One draws complete designs such as app screens, infographics and posters, with the text spelled correctly and, if you ask, a transparent background. The other takes a flat design image and cuts it into separate transparent layers you can edit one by one. Both are free to use commercially under the MIT license, but they need a large graphics card.

Summary

inclusionAI, Ant Group's open AI initiative, published Ming-Image 0.1 Design in September 2026 as two models under the MIT license. Ming-Image-0.1-Design generates complete visual designs, such as user interfaces, infographics, posters and other text-heavy compositions, and it can output transparent RGBA images. Ming-Image-0.1-Design-Layer goes the other way and splits a flattened design image into independent transparent layers. The card's 6B figure refers to the diffusion transformer. Each repository also holds a multimodal language model of about 34 GB and a 6.2 GB connector, so the downloads are 52.9 GB for the text-to-image model and 65.2 GB for the Layer model.

The default text-to-image setting is 12 steps without guidance at 2,048 by 2,048 pixels, and sizes within a ratio of 1 to 4 up to 4 to 1 are accepted. For better results the README recommends a Figma-style structured prompt, a JSON caption with back-to-front layers, exact coordinates and color specs, which a separate language model such as Ling-3.0-flash-VL or Qwen3.8-27B writes from your short request, and the system prompt is published. The Layer model reads a per-layer plan that a second prompt enhancer helps produce, and it returns the requested number of layers as standalone images. The README calls a single GPU with at least 80 GiB of memory, running BF16, the default and minimum validated setup, and points to vLLM-Omni recipes for serving. inclusionAI also links two agent skills. Ling UI Design uses the generated references and layer decomposition to help an agent build UI code, and Image to Editable PPT rebuilds a generated slide image as an editable PowerPoint slide.

The card's headline claim is a placement on the Artificial Analysis UI/UX Design leaderboard, shown as a chart whose numbers we could not read, and the README gives no benchmark scores in text. The 0.1 version number signals an early release. In the 30 days to 11 November 2026, Hugging Face counted about 1,800 downloads of the Design model and about 600 of the Layer model.

Good for

  • Mockups of app screens, infographics, posters and other designs full of text, including assets with a transparent background
  • Turning a flat design image into separate, editable transparent layers
  • Agent workflows that build UI code or an editable slide from a generated visual
  • Commercial work, since the MIT license has no use restrictions of its own

Watch out for

  • The 6B figure covers only the diffusion transformer, the repositories are 52.9 GB and 65.2 GB, and the validated setup is one GPU with at least 80 GiB
  • Best results need a separate language model to rewrite your request as structured JSON, or as a layer plan for the Layer model
  • The README shows its leaderboard result only as a chart, with no scores in text
  • It is a 0.1 release, and the card documents no image editing apart from layer decomposition
  • The two models and the agent skills are separate pieces that you set up yourself

Performance claims come from the developer and have not been independently verified.

How much memory does it need?

Memory figures for each variant of Ming-Image 0.1 Design. Image, video, speech and music pipelines also load a text encoder and a decoder, so the memory needed can be larger than the main checkpoint.

Variants Parameters Download size Memory needed Fits on
Ming-Image-0.1-Design (text-to-image) 6.2B 53 GB 80 GB Workstation
Ming-Image-0.1-Design-Layer (layer decomposition) 6.2B 65 GB 80 GB Workstation

Download size covers every component needed to run the model, such as the generator, text encoder and decoder. Memory needed is a stated or published measured minimum where one exists; otherwise it is the download size plus about 15%. Larger images, longer clips and higher resolutions need more. The developer states a minimum of 80 GB of GPU memory.

The license in plain words

The shortest and simplest permissive license. You can do almost anything with the model as long as you keep the copyright and license notice.

More about MIT

See also

Frequently asked questions

What is Ming-Image 0.1 Design?
Ming-Image 0.1 Design is a pair of image models for designers. One draws complete designs such as app screens, infographics and posters, with the text spelled correctly and, if you ask, a transparent background. The other takes a flat design image and cuts it into separate transparent layers you can edit one by one. Both are free to use commercially under the MIT license, but they need a large graphics card.
Can I use Ming-Image 0.1 Design commercially?
Yes. Ming-Image 0.1 Design is released under MIT, a permissive license. You can use it commercially if you follow the standard conditions, such as keeping the license notice.
Is Ming-Image 0.1 Design open source?
Ming-Image 0.1 Design is released under MIT, a license approved by the Open Source Initiative. Training data and training code may not be public, so it is best described as open weights under an open-source license.
Can I run Ming-Image 0.1 Design on my own computer?
Only on a workstation. The smallest variant, Ming-Image-0.1-Design (text-to-image), needs about 80 GB of memory at 4-bit precision, so plan for a 48 to 96 GB GPU, several graphics cards, or a Mac with plenty of memory.
When was Ming-Image 0.1 Design released?
Ming-Image 0.1 Design was released by inclusionAI (Ant Group) on September 17, 2026.

Official links

Sources

Specifications come from these primary sources. Benchmark claims, where mentioned, are reported by the developer.

Release date and facts last verified November 11, 2026.

Related models

Model Released Parameters Context License Smallest variant needs
Qwen-Image-2.1 Alibaba Qwen 7.1B – Qwen Research License (Non-commercial) 20 GB Gaming GPU
HiDream-O1-Image HiDream.ai 8.8B – MIT (Permissive) 20 GB Gaming GPU
Ideogram 4 Ideogram 9.3B – Ideogram Non-Commercial (Non-commercial) 19 GB Gaming GPU
Krea 2 Krea 12.8B – Krea 2 Community (Conditional) 41 GB Workstation