ByteDance-Seed/Bagel
A ByteDance multimodal foundation model that unifies understanding and generation across text, images, and other modalities.

Velocity · 7d
+14
★ / day
Trend
→steady
star history
Bagel is an open-source unified multimodal model developed by ByteDance Seed. It combines multimodal understanding and generation capabilities in a single foundation model, supporting text, images, and other modalities. The project provides model weights on HuggingFace, training code, and a demo interface for the 7B-scale MoT (Mixture-of-Transformers) variant.