ERNIE-Image is an open text-to-image generation model developed by the ERNIE-Image team at Baidu. It is built on a single-stream Diffusion Transformer (DiT), with only 8B DiT parameters, it reaches state-of-the-art performance among open-weight text-to-image models.
Puffin Series: Towards Unified Multimodal 3D World Models **Puffin** is a series of unified multimodal models advancing toward 3D world modeling. 04: The paper of **Puffin-World** is released on arXiv. 23: The model weights of **Puffin-World** (Base / Pro / Caption) are released on Hugging Face. 23: The scripts of the dataset construction pipeline have been released. 23: The camera captions (by our method) of the commonly used large-scale text-to-image datasets, such as megalith-10m, have been
released. 22: The training and evaluation code of **Puffin-World** is released. 26: Puffin has been accepted at ICLR 2026. 15: Puffin-4M dataset reached 20,000 downloads on Hugging Face within three months of release. 10: The scripts of the camera-centric evaluation has been released. 10: The paper, project page, code, model, dataset, and demo of Puffin are online. 08673 (ICLR 2026) | Page | 🤗 KangLiao/Puffin | 🤗 Puffin-4M | | | **Puffin-World** *Native 3D World States* | arXiv:2609. 04196 | Page | 🤗 KangLiao
/Puffin-World | 🤗 Puffin-16M | | 🗞️ License This project is licensed under NTU S-Lab License 1. 📚 Citation If you find Puffin useful for your research or applications, please cite our papers using the following BibTeX:
Frequently asked questions
How many stars and forks are recorded?
KangLiao929/Puffin has 515 stars and 23 forks.
What is the repository rank?
KangLiao929/Puffin is ranked 17,962.
What language and category are recorded?
KangLiao929/Puffin has 515 stars, uses Python and belongs to Models.
Which repository is closest in rank?
KangLiao929/Puffin is compared with Atten4Vis/LW-DETR, ranked 17,980 with 514 stars.
README
Puffin Series: Towards Unified Multimodal 3D World Models
Puffin is a series of unified multimodal models advancing toward 3D world
modeling. It starts from camera-centric spatial intelligence — understanding
and generating the world from arbitrary viewpoints and orientations — and
scales to a world model that represents worlds through three complementary
native 3D world states (physics, geometry, and appearance),
supporting camera-to-world understanding, camera-controllable generation, and
image-/text-to-3D world generation without external perception or
reconstruction modules. Each release lives in its own subdirectory of this
repository:
Puffin/ — Thinking with Camera: A Unified Multimodal Model
for Camera-Centric Understanding and Generation (ICLR 2026)
Puffin-World/ — Puffin-World: Scaling a Unified
Multimodal Model with Native 3D World States (2026)
📝 Changelog & News
[x] 2026.09.04: The paper of Puffin-World is released on arXiv.
[x] 2026.08.23: The model weights of Puffin-World (Base / Pro / Caption) are released on Hugging Face.
[x] 2026.08.23: The scripts of the dataset construction pipeline have been released.
[x] 2026.08.23: The camera captions (by our method) of the commonly used large-scale text-to-image datasets, such as megalith-10m, have been released.
[x] 2026.08.22: The training and evaluation code of Puffin-World is released.
[x] 2026.01.26: Puffin has been accepted at ICLR 2026.
[x] 2026.01.15: Puffin-4M dataset reached 20,000 downloads on Hugging Face within three months of release.
[x] 2026.01.10: The scripts of the camera-centric evaluation has been released.
[x] 2025.10.10: The paper, project page, code, model, dataset, and demo of Puffin are online.
If you find Puffin useful for your research or applications, please cite our
papers using the following BibTeX:
@article{liao2025puffin,
title={Thinking with Camera: A Unified Multimodal Model for Camera-Centric Understanding and Generation},
author={Liao, Kang and Wu, Size and Wu, Zhonghua and Jin, Linyi and Wang, Chao and Wang, Yikai and Wang, Fei and Li, Wei and Loy, Chen Change},
journal={arXiv preprint arXiv:2510.08673},
year={2025}
}
@article{liao2026puffinworld,
title={Puffin-World: Scaling a Unified Multimodal Model with Native 3D World States},
author={Liao, Kang and Luo, Yihang and Wu, Xiao-Ming and Jin, Linyi and Wu, Size and Lin, Chunyu and Zhao, Yao and Wang, Fei and Li, Wei and Loy, Chen Change},
journal={arXiv preprint arXiv:2609.04196},
year={2026}
}
Puffin Series: Towards Unified Multimodal 3D World Models
