Skip to content
Trending storyDeveloping

ModelScope releases MOSS-Image-Base, an 8.56B open-weight text-to-image model

1 article1 sourcesince Oct 16Last article 1h ago ·

Overview

AISummary of 1 article

ModelScope releases MOSS-Image-Base, an 8.56B-parameter open-weight text-to-image foundation model, under an Apache 2.0 license for the model weights and inference code.

Per the announcement on X, the model accepts descriptive prompts, tag sequences, or JSON layouts that specify text and object placement, including bounding boxes.

ModelScope describes the model as a base checkpoint for fine-tuning. It ships with Diffusers support, a Gradio interface, CPU offloading options, and a native ComfyUI plugin.

Written by AI from the articles below · updated Oct 16, 6:17 AM ET

Check the sources:

Latest developmentOct 16, 6:00 AM ET
MOSS-Image-Base: an 8.56B open-weight text-to-image model with layout control

Article timeline

The articles in this story. Times are ET.

Oct 16
  1. ModelScopeOfficial
    MOSS-Image-Base: an 8.56B open-weight text-to-image model with layout control

    AIModelScope releases MOSS-Image-Base, an 8.56B-parameter open-weight text-to-image foundation model under Apache 2.0. It accepts descriptive prompts, tag sequences, or JSON layouts that specify text and object placement, including bounding boxes. The model is designed as a base checkpoint for fine-tuning and includes Diffusers support, a Gradio interface, CPU offloading options, and a native ComfyUI plugin.

    Image from @ModelScope2022's post

Heat trend

Not enough continuous observations to show a trend yet.