MiniMax releases MiniMax-H3-Image-PE prompt expansion model
MiniMaxAI released the 27B MiniMax-H3-Image-PE model for expanding short text-to-image requests into detailed captions and suggesting aspect ratios. The FP4 checkpoint is derived from Qwen3.8-27B and includes SGLang serving instructions, task and system prompts, a client example, and local Diffusers integration. The repository also provides an image-editing prompt template and message-assembly example, but that workflow remains under development.
The release gives developers an open local workflow for preparing prompts and aspect ratios for MiniMax-H3-Image, with deployment guidance for SGLang and Diffusers.
MiniMaxAI/MiniMax-H3-Image-PE Text Generation • 27B • Updated 10 days ago • 4.19k • 133
pipeline_tag: text-generation license: other license_name: minimax-h3-image-community-license-agreement license_link: https://huggingface.co/MiniMaxAI/MiniMax-H3-Image/blob/main/LICENSE library_name: transformers language:
- en tags:
- prompt-expansion
- text-to-image
- minimax-h3
- fp4
- sglang base_model:
- Qwen/Qwen3.8-27B

#MiniMax#MiniMax-H3-Image-PE#MiniMax-H3-Image#Qwen3.8-27B#SGLang#Diffusers#XGrammar#Transformers