MiniMax (MiniMaxAI) · Open 16MP Image Model

MiniMax-H3-Image

MiniMax-H3-Image is an open-weight diffusion model for text-to-image generation and multi-reference image editing, part of the MiniMax H3 family. The page says it generates about 16-megapixel images at any aspect ratio (up to 4096x4096) without an upscaler, and that it is distilled to 6 sampling steps with no classifier-free guidance. Its 33.1B-parameter diffusion transformer supports up to 16 reference images, and the page describes it as having the strongest instruction following among open-source models. It is available through MiniMax's API and app, and on Hugging Face under the MiniMax-H3-Image Community Licence, with support listed for SGLang, vLLM-Omni, diffusers and ComfyUI.