LLaDA-Image

Ant Group (inclusionAI)🇨🇳 CNMajor lab· Aug 28, 2026

Imageopen weights

LLaDA-Image generates images from text and edits existing images from instructions, using one checkpoint without a separate editing backbone. The modelcard reports 53.53 in English and 53.38 in Chinese on Qwen-Image-Bench.

Summary of the model card

License: apache-2.0
Modalities: text → image

2 variants

FP8				HF
Standard				HF

Sources

HuggingFace Paper

Deutsch

LLaDA-Image erzeugt Bilder aus Text und bearbeitet vorhandene Bilder anhand von Anweisungen, ohne separates Editing-Modell. Laut Modelcard erreicht es 53.53 (Englisch) und 53.38 (Chinesisch) im Qwen-Image-Bench.

Provenance

Detected: Sep 27, 2026
First source: huggingface
Description: Summary of the model card
Editorially reviewed: —