MiniMax Open-Sources MiniMax-H3-Image Model, Capable of Directly Generating 16-Megapixel Images

November 10 — According to Hugging Face, MiniMax has open-sourced its image generation and editing model MiniMax-H3-Image. A single set of weights supports both text-to-image and image editing, with up to 16 reference images as input. The model can directly generate images of about 16 megapixels without an additional super-resolution model, with resolutions ranging from 4096×4096 to 6272×2688, and sampling steps distilled down to 6. Its diffusion Transformer has 33.1 billion parameters, with weights of about 68.65GB, and can be deployed using SGLang, vLLM-Omni, diffusers and ComfyUI. MiniMax says the model has the strongest instruction-following capability among open-source models. The model is also available on the MiniMax API and Hailuo AI, and is released under a community license agreement.