NEO

NEO Series: Native Vision-Language Models

• 2026/11: From Pixels to Actions -- Towards Native Vision-Language-Action Policies (Arxiv 2026)

• 2026/04: SenseNova-U1 with NEO-unify Architecture (Technical Report 2026)

• 2026/02: NEO-unify: Building Native Multimodal Unified Models End to End (HuggingFace Blog 2026)

• 2025/12: From Pixels to Words -- Towards Native One-Vision Models at Scale (Arxiv 2026)

• 2025/09: From Pixels to Words -- Towards Native Vision-Language Primitives at Scale (ICLR 2026)

📜 News

[2026/11] 🔥 The paper, project page, weights, dataset, and training code of NEO-VLA are released !
[2026/05] 🔥 The paper, weights, and evaluation code of NEO-ov are released !
[2025/12] 💥 NEO-ov has been completed !
[2026/01] 🔥 The training code of NEO is released !
[2025/10] 🔥 The paper, weights, and evaluation code of NEO are released !
[2025/09] 💥 NEO has been completed !

📋 Todo List

• Evaluation guide (NEO-ov)
• Evaluation guide (NEO)
• Training guide
• VLA training guide (NEO-VLA)

🤖 Model Zoo

We release 2B and 9B NEO-series models across different training stages, and the 2B NEO-VLA policy with its action pretraining dataset.

💡 Motivation

• What constraints set native VLMs apart from modular ones, and to what extent can they be overcome?

• How to make native VLMs more accessible and democratized, thereby accelerating their progress?

💡 Highlights

• 🔥 Native Architecture: NEO innovates a native VLM primitive that unifies pixel-word encoding, alignment, and reasoning within an encoder-free dense model architecture.

• 🔥 Superior Efficiency: Using under 10% of image-text data, NEO develops strong visual perception from scratch, rivaling top-tier modular VLMs and outperforming native encoder-free ones.

• 🔥 Promising Roadmap: NEO pioneers a promising route for scalable and powerful native encoder-free VLMs, paired with diverse reusable components that foster a cost-effective and extensible ecosystem.

🚀 Performance

🧠 Single-Image Understanding

🎬 Multi-Image & Video Understanding

🌐 Spatial Intelligence

✒️ Citation

If NEO series is helpful for your research, please consider star ⭐ and citation 📝 :

@article{Diao2025NEO,
  title        = {From Pixels to Words--Towards Native Vision-Language Primitives at Scale},
  author       = {Diao, Haiwen and Li, Mingxuan and Wu, Silei and Dai, Linjun and Wang, Xiaohua and Deng, Hanming and Lu, Lewei and Lin, Dahua and Liu, Ziwei},
  journal      = {arXiv preprint arXiv:2510.14979},
  year         = {2025}
}

@article{Diao2026NEOov,
  title        = {From Pixels to Words--Towards Native One-Vision Models at Scale},
  author       = {Diao, Haiwen and Wang, Jiahao and Wu, Penghao and Dong, Yuhao and Niu, Yuwei and Zhu, Yue and Cai, Zhongang and Fan, Weichen and Dai, Linjun and Wu, Silei and others},
  journal      = {arXiv preprint arXiv:2605.28820},
  year         = {2026}
}

@misc{sensenova2026neounify,
  title        = {NEO-unify: Building Native Multimodal Unified Models End to End},
  author       = {SenseNova},
  journal      = {Hugging Face blog},
  url          = {https://huggingface.co/blog/sensenova/neo-unify},
  year         = {2026}
}

@article{sensenova2026sensenovau1,
  title        = {SenseNova-U1: Unifying Multimodal Understanding and Generation with NEO-unify Architecture},
  author       = {Diao, Haiwen and Wu, Penghao and Deng, Hanming and Wang, Jiahao and Bai, Shihao and Wu, Silei and Fan, Weichen and Ye, Wenjie and Tong, Wenwen and Fan, Xiangyu and others},
  journal      = {arXiv preprint arXiv:2605.12500},
  year         = {2026}
}

@article{Fan2026NEOVLA,
  title        = {From Pixels to Actions--Towards Native Vision-Language-Action Policies},
  author       = {Fan, Weichen and Diao, Haiwen and Wu, Penghao and Hu, Siyuan and Huang, Ziqi and Liu, Tianqi and Sun, Mingze and Wang, Ziyi and Cai, Zhongang and Deng, Hanming and Wang, Quan and Yang, Lei and Lu, Lewei and Hong, Fangzhou and Chen, Zhaoxi and Lin, Dahua and Liu, Ziwei},
  journal      = {arXiv preprint arXiv:2611.06918},
  year         = {2026}
}

📄 License

The content of this project itself is licensed under LICENSE.