Lumina Dimoo is a unified discrete diffusion model for multimodal generation and understanding. It supports text-to-image, image-to-image, and image understanding with efficient sampling and strong benchmark results.