Oriel — AI & technology worth noticing, every day.

MiniMaxAI/MiniMax-H3-Image

AI model · https://huggingface.co/MiniMaxAI/MiniMax-H3-Image

First seen by Oriel on .

AI summary

MiniMax-H3-Image is an open-source image generation and editing model that produces roughly 16-megapixel images in six sampling steps while emphasizing instruction following, realism, and controllability.

  • One checkpoint supports text-to-image generation and editing with up to 16 reference images, across aspect ratios from 4096x4096 to 6272x2688.
  • Its 33.1B-parameter, 50-layer diffusion transformer uses MiniMax-H3's unchanged Qwen3-VL text encoder and a VAE with 16x spatial downsampling.
  • The model uses six inference steps without classifier-free guidance; other step counts, guidance scales beyond 1, and negative prompts are rejected.
  • Local deployment supports SGLang, vLLM-Omni, diffusers, and ComfyUI, with sequence parallelism available across multiple NVIDIA or AMD GPUs.
  • The transformer download is 66.28 GB, while the separately loaded MiniMax-H3 text encoder adds 66.7 GB; usage is governed by the Community License Agreement.

#image generation #image editing #diffusion transformer #multimodal models

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Likes

Total likes reported by the source.

  • Hugging Face Models: 1,348, latest observation ; +1,190 over observed period since

Downloads (30d)

Rolling 30-day downloads. A change in this value is not new downloads during the observed period.

  • Hugging Face Models: 609,130, latest observation ; +608,463 over observed period since

Aliases

SourceExternal IDURL
Hugging Face ModelsMiniMaxAI/MiniMax-H3-Imagehttps://huggingface.co/MiniMaxAI/MiniMax-H3-Image Read discussion