Ant Group · image_gen
Ming-Image-0.1-Design-Layer
Ming-Image-0.1-Design-Layer decomposes a flattened design image into RGBA layers using an image and a layer plan. This MIT-licensed checkpoint serves a different task from the Design text-to-image model.
Downloadable checkpoint · MIT
Official sources reviewed 2026-10-04. This page documents access and specifications; it does not report a model benchmark.
Task and deployment contract
Download the Design-Layer checkpoint and run the layer-decompose task with an input image and a layer plan.
Provide an input image and either a detailed layer specification or a requested layer count. When a prompt declares the layer count, that plan controls the number of outputs.
The official example decomposes a card design into six transparent PNG files. Check each alpha channel and composite the outputs to assess whether the original design is preserved.
Reference settings are BF16, 12 sampling steps and CFG 2.0, validated on one CUDA GPU with 80 GiB VRAM. The model card does not establish lower hardware requirements or production throughput.
This is a layer-decomposition checkpoint, not a general text-to-image endpoint. MIT licensing of the model does not grant rights to every input design.
Billing and limits
No current provider rate for this exact configuration is recorded here. Open weights still require compute and operations; an unlisted rate does not mean a hosted API is free.
The documented working-resolution buckets are 512 and 1024. Output preserves the input aspect ratio; a chat-token context window is not applicable.
Fast Pivot — Access hundreds of AI models from a single API — competitive pricing, reliable uptime, enterprise-grade infrastructure.