Gemini Robotics: Bringing AI into the Physical World

• DOI: 10.48550/arxiv.2503.20020
• Publication Date: 2025-03-25
• Authors: Gemini Team; Saminda Abeyruwan; Joshua Ainslie; Jean-Baptiste Ala...; Montserrat Gonzal...; Travis Armstrong; Ashwin Balakrishna; Robert Baruch; Maria Bauzá; Michiel Blokzijl; Steven Bohez; Konstantinos Bous...; Anthony Brohan; Thomas Buschmann; Arunkumar Byravan; Serkan Cabi; Ken Caluwaerts; Federico Casarini; Oscar Chang; José Enrique Chen; Xi Chen; Hao-Tien Lewis Ch...; Krzysztof Choroma...; David B. D’Ambrosio; Sudeep Dasari; Todor Davchev; Coline Devin; Norman Di Palo; Tianli Ding; Adil Dostmohamed; Danny Driess; Yilun Du; Debidatta Dwibedi; Michael Elabd; Claudio Fantacci; Cody Fong; Erik Frey; Chuyuan Fu; Marissa Giustina; Keerthana Gopalak...; Laura Graesser; Leonard Hasenclever; Nicolas Heess; Brandon Hernaez; Alexander Herzog; Roswitha Hofer; Jan Humplik; Atıl Işçen; Mithun George Jacob; Deepali Jain; Ryan R. Julian; Dmitry Kalashnikov; Mustafa Emre Kara...; Stefani Karp; Chase Kew; J. Russell Kirkland; Sean Kirmani; Yuheng Kuang; Thomas Lampe; Antoine Laurens; Isabel Leal; Alex X. Lee; Tsang-Wei Edward Lee; Jacky Liang; Yixin Lin; Sharath Maddineni; Anirudha Majumdar; Assaf Hurwitz Mic...; Robert Moreno; Michael Neunert; Francesco Nori; Carolina Parada; Emilio Parisotto; Peter Pástor; Acorn Pooley; Kanishka Rao; Krista Reymann; Dorsa Sadigh; Stefano Saliceti; Pannag Sanketi; Pierre Sermanet; Dhruv Shah; Mohit Sharma; Kevin M. Shea; Chaoyang Shu; Vikas Sindhwani; Sumeet Singh; Radu Soricut; Jost Tobias Sprin...; Rachel Sterneck; Razvan Surdulescu; Jie Tan; Jonathan Tompson; Vincent Vanhoucke; Jake Varley; Grace Vesom; Giulia Vezzani; Oriol Vinyals; Ayzaan Wahid; Stefan Welker; Paul Wohlhart; Fei Xia; Ted Xiao; Annie Xie; Jinyu Xie; Peng Xu; Sichun Xu; Ying Xu; Zhuo Xu; Yuxiang Yang; Rui Yao; Sergey Yaroshenko; Wenhao Yu; Wentao Yuan; Jingwei Zhang; Tingnan Zhang; Allan Zhou; Yuxiang Zhou
• Keywords: FOS: Computer and information sciences; Computer Science - Robotics; Robotics (cs.RO)

Recent advancements in large multimodal models have led to the emergence of remarkable generalist capabilities digital domains, yet their translation physical agents such as robots remains a significant challenge. This report introduces new family AI purposefully designed for robotics and built upon foundation Gemini 2.0. We present Robotics, an advanced Vision-Language-Action (VLA) model capable directly controlling robots. Robotics executes smooth reactive movements tackle wide range complex manipulation tasks while also being robust variations object types positions, handling unseen environments well following diverse, open vocabulary instructions. show that with additional fine-tuning, can be specialized including solving long-horizon, highly dexterous tasks, learning short-horizon from few 100 demonstrations adapting completely novel robot embodiments. is made possible because builds on top Robotics-ER model, second we introduce this work. (Embodied Reasoning) extends Gemini's reasoning into world, enhanced spatial temporal understanding. enables relevant detection, pointing, trajectory grasp prediction, multi-view correspondence 3D bounding box predictions. how combination support variety applications. discuss address important safety considerations related class models. The marks substantial step towards developing general-purpose realizes AI's potential world.