[MiniMax Open-Sources MiniMax-H3-Image Model, Capable of Generating 16-Megapixel Images] According to Hugging Face, MiniMax has open-sourced its image generation and editing model, MiniMax-H3-Image. The same set of weights supports both text-based and image editing, and can accept up to 16 reference images as input. This model does not require a separate super-resolution model and can directly generate images of approximately 16 megapixels, covering resolutions from 4096×4096 to 6272×2688, with a sampling step count reduced to 6 steps. Its diffusion Transformer has 33.1 billion parameters and approximately 68.65GB of weights, and can be deployed with SGLang, vLLM-Omni, diffusers, and ComfyUI. MiniMax claims that this model has the strongest command compliance among open-source models.