MiniMax releases open-weight MiniMax-H3-Image for generation and editing
Original title: MiniMaxAI/MiniMax-H3-Image — new model trending #27 on Hugging Face
The checkpoint outputs about 16-megapixel images in 6 steps and accepts up to 16 reference images.
MiniMax published MiniMax-H3-Image, an open-weight model from the MiniMax H3 family for text-to-image generation and image editing in one checkpoint. It is listed as trending on Hugging Face and is also offered through MiniMax APIs and the Hailuo apps. The card says the model follows the H3 diffusion transformer without audio layers and keeps the MiniMax-H3 text encoder.
Unconfirmed: The card's claim of the strongest instruction following among open-source models is MiniMax's own and is not independently verified here.
- The main checkpoint is BF16 and listed at 68.65 GB.
- Optional MiniMax-H3-Image-PE is an NVFP4 prompt expander at 17.9 GB.
- Outputs span about 4096x4096 to 6272x2688 without an upscaler.
- Sampling is distilled to 6 steps with no classifier-free guidance.
- Vendors
- MiniMax
- AI models
- MiniMax-H3-ImageMiniMax-H3-Image-PE
From the original article
9 min read at huggingface.co<img width="100%" src="assets/minimax-h3-image. jpg" alt="MiniMax H3 image, Open Weights"> <a href="https://hailuoai. video" target="_blank"><img src="https://img.
By MiniMaxAI · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.