MOSS Image Base vs. SFT
Two checkpoints in one model family. Choose a foundation for local customization, or a fine-tuned model for text-to-image generation.
What is MOSS Image Base?
Base is the foundation text-to-image checkpoint from OpenMOSS. It accepts natural-language prompts, tag sequences, and structured JSON layouts. It is also a starting point for task-specific fine-tuning and LoRA training.
What is MOSS Image SFT?
SFT is supervised fine-tuned from Base on curated data. It shares the architecture and inference pipeline with Base. The official comparison shows clearer text grouping and a more organized whiteboard layout in its example; it is not a guarantee that every prompt will perform better.
Which model does this generator use?
MOSS Studio is configured for MOSS Image SFT. Base is described here for people exploring the model family; it is not an additional online model option in the current generator.