Support the model Bagel

### Feature Idea

https://github.com/ByteDance-Seed/Bagel

An open-source multimodal foundation model with 7B active parameters (14B in total), trained on large-scale interleaved multimodal data. BAGEL outperforms current top open-source VLMs such as Qwen2.5-VL and InternVL-2.5 on the standard multimodal understanding leaderboard, and offers text-to-image quality comparable to powerful professional generators like SD3. Additionally, BAGEL demonstrates qualitative results superior to leading open-source models in classic image editing scenarios. More importantly, it extends to free-form visual composition, multi-view synthesis, and world navigation, features that constitute "world modeling" tasks beyond the scope of previous image editing models.


### Existing Solutions

_No response_

### Other

_No response_