MiniMax has open-sourced MiniMax H3 Image, a native 4K direct-output image model capable of generating images in around 10 seconds and supporting up to 16 reference images, ranking first on the open-weight image model leaderboard. Built upon the omni-modal video model H3, this model excels in photorealistic rendering, text generation, face preservation, image-to-image secondary creation, understanding of physical world structure, and complex instruction following—covering scenarios ranging from everyday use to professional production. Currently, an active global derivative ecosystem has formed around MiniMax's open-source models. By carving out an ecological niche through open-weight releases, MiniMax has built a commercial loop through hosted APIs, subscription plans, and commercial licensing models.
