InclusionAI's LLaDA-Image-FP8 supports creating an image from text and editing a supplied reference image with the same checkpoint. The model card describes a shared pipeline: generation takes a prompt, while editing adds a reference image and an instruction. This gives developers one model family for both operations, with a recommended 50-step sampling configuration for the base variant.
Terms in this piece · Glossary
distillation — Training a small, cheap model to imitate a big one's outputs, keeping much of the capability at a fraction of the cost.
