LLaDA-Image ComfyUI AIO conversions by T8, 2026.

Original model and weights: inclusionAI/LLaDA-Image and inclusionAI/LLaDA-Image-Turbo.
Original project: https://github.com/inclusionAI/LLaDA-Image
Pinned source model cards declare Apache-2.0:
Base: e4e2703f410f7ddb6ee8d6b09dac6a8ec5093039
Turbo: f4afc52d925bbac4e22a1c947111fc1f127e37e5

Changes: original component tensors are repackaged under ComfyUI AIO tensor
namespaces; component configuration and tokenizer data are embedded into one
safetensors checkpoint. No retraining or quantization is performed.
Conversion tooling: https://github.com/T8mars/Comfyui-LLaDa-Image-T8
Detailed source-file hashes and output hashes accompany each checkpoint.

Model weights use Apache-2.0, retained from the original model cards.
The separate ComfyUI node integration code has its own license in the code repo.