InclusionAI releases LLaDA-Image: one 6B diffusion DiT for generation and editing

bdsqlsz · x · 2026-09-04

TheInclusionAI introduced LLaDA-Image and LLaDA-Image-Turbo. A single 6B DiT unifies text-to-image generation and instruction-guided editing, with both backbone and image-gen trained as diffusion models in one framework. Outputs cover photorealistic images, posters, ads, and bilingual typography; the Turbo variant cuts inference to 2-4 steps.

Related event: Ant's inclusionAI open-sources 6B image generation-editing model LLaDA-Image(3 posts)→

Original post →

More from Multimodal

Multimodal channel →