Archive · 27 May 2025
ByteDance Unveils Open-Source Bagel AI Model
Imagine a computer program that can understand pictures and words, like a super-smart artist.
ByteDance created one called Bagel.
It's an AI that can look at images and text together to understand what's happening.
Bagel can create, change, and edit pictures, making it easy to add things, remove stuff, or even change how an image looks.
You can try it out online, and it's free for anyone to use for learning or in businesses.
The creators say it works better than some other similar programs, and it knows how the world works visually to make better edits.
ByteDance has released Bagel, a new open-source multimodal AI model.
Bagel is a visual language model (VLM) capable of understanding, generating, and editing images.
The model has 14 billion parameters, with 7 billion active at a time, and was trained on interleaved multimodal data.
ByteDance claims Bagel outperforms other open-source VLMs in image editing.
The model is available on GitHub and Hugging Face for both academic and commercial use.
- Who
- ByteDance
- What
- Released Bagel, a multimodal AI model.
- When
- Last week
- Where
- Beijing
- Why
- To provide a more efficient and capable image editing tool using multimodal data and understanding
- Type
- Multimodal AI Model
- Total Parameters
- 14 billion
- Active Parameters
- 7 billion
- Capabilities
- Image generation, editing, and understanding
- License
- Open source
- Availability
- GitHub and Hugging Face