titleGigaBrain-0.7: Scaling Embodied Foundation Models to Emergent Capabilities with a Three-System Architecture
authorsGigaBrain Team, Angen Ye, Axiang Sun, Can Jin, Chenxi Cheng, Chong Shi, Dengke Shang, Dingqian Zhang, Guan Huang, Guangqiang Wang, Guangqing Ding, Guo Li, Hangcong Li, Hengyu Zhong, Hongtao Lu, Jianbo Qin, Jiming Mao, Jing Zhu, Jindi Lv, Jingzhi Cui, Junjie Xie, Junyi Bao, Kai Liu, Lei Yuan, Limin Long, Lv Feng, Mingming Yu, Peng Li, Pengfei Yi, Qi Li, Qianli Zhang, Qingfang Li, Qitang Hu, Rui Zhang, Shaoyan Sun, Shibo Sun, Shiying Duan, Tenghui Chen, Tianze Liu, Weijie Ke, Wenyao Xue, Xiaofeng Wang, Xiaoyu Tian, Xinyu Liu, Xinze Chen, Yang Wang, Yankai Wang, Yejun Zeng, Yifan Li, Yifei Nie, Yilong Li, Yilong Liu, Yongchao Feng, Yumeng Wang, Yun Ye, Zhichao Liu, Ziheng He, Zonghai Yang, Zheng Zhu
date2026-08-16
venuearXiv (preprint, not peer reviewed)
idarXiv:2608.15875
affiliationGigaAI
upvotes27 (Hugging Face Daily Papers, 2026-11-01)
linksabs pdf hf code
primerVision-language-action models
abstractVision-language-action (VLA) models have become a dominant paradigm for generalist embodied agents, demonstrating strong complex and long-horizon task completion in structured settings. Yet it remains an open question whether current VLA systems can benefit from more effective architectural design, scale to substantially larger and more heterogeneous data regimes, and achieve broader generalization across tasks and embodiments. To this end, we present GigaBrain-0.7, an embodied foundation model with substantially improved generalization across diverse robot embodiments. Specifically, GigaBrain-0.7 unifies understanding, prediction, and action through a three-system architecture, scales pretraining to over 37,000 hours of heterogeneous embodied data, and introduces one-stage alignment training that jointly optimizes vision-language understanding and multi-embodiment action generation. Compared with the preceding GigaBrain-0 series and prior state-of-the-art models including π_{0.5}, GigaBrain-0.7 achieves substantial improvements in foundation zero-shot capabilities, language-conditioned instruction following, and post-training task success rates. In particular, on our in-house Maker H01 platform and mainstream robot embodiments, GigaBrain-0.7 demonstrates strong task adaptability and completion ability across both home and industrial scenarios. All training code and pretrained model weights will be released.

cite

@misc{team2026gigabrain,
  title={GigaBrain-0.7: Scaling Embodied Foundation Models to Emergent Capabilities with a Three-System Architecture},
  author={GigaBrain Team and Angen Ye and Axiang Sun and Can Jin and Chenxi Cheng and Chong Shi and Dengke Shang and Dingqian Zhang and Guan Huang and Guangqiang Wang and Guangqing Ding and Guo Li and Hangcong Li and Hengyu Zhong and Hongtao Lu and Jianbo Qin and Jiming Mao and Jing Zhu and Jindi Lv and Jingzhi Cui and Junjie Xie and Junyi Bao and Kai Liu and Lei Yuan and Limin Long and Lv Feng and Mingming Yu and Peng Li and Pengfei Yi and Qi Li and Qianli Zhang and Qingfang Li and Qitang Hu and Rui Zhang and Shaoyan Sun and Shibo Sun and Shiying Duan and Tenghui Chen and Tianze Liu and Weijie Ke and Wenyao Xue and Xiaofeng Wang and Xiaoyu Tian and Xinyu Liu and Xinze Chen and Yang Wang and Yankai Wang and Yejun Zeng and Yifan Li and Yifei Nie and Yilong Li and Yilong Liu and Yongchao Feng and Yumeng Wang and Yun Ye and Zhichao Liu and Ziheng He and Zonghai Yang and Zheng Zhu},
  year={2026},
  eprint={2608.15875},
  archivePrefix={arXiv},
  url={https://arxiv.org/abs/2608.15875}
}