Google DeepMind has launched two new AI models, Gemini Robotics and Gemini Robotics-ER, tailored for robotics applications.
Built on the Gemini 2.0 model, these models introduce advanced capabilities for controlling robots and understanding spatial environments.
Go deeper with GlobalData
Gemini Robotics is an advanced vision-language-action model, incorporating physical actions as a new output modality to directly control robots.
This model is designed to enhance the interaction between robots and their environments, enabling them to perform a wider range of real-world tasks, the company said.
The Gemini Robotics-ER model focuses on advanced spatial understanding, allowing roboticists to utilise Gemini’s embodied reasoning abilities.
This model enables robots to have an advanced understanding of their surroundings, facilitating integration with existing low-level controllers.
In a blog post, Google revealed that the Gemini Robotics model was trained on data from the bi-arm robotic platform, ALOHA 2, demonstrating its capability to control a bi-arm platform based on Franka arms.
This model can be specialised for complex embodiments, such as the humanoid Apollo robot developed by Apptronik.
Google noted that the company plans to partner with Apptronik to develop the next generation of humanoid robots using Gemini 2.0.
“We’re also working with a selected number of trusted testers to guide the future of Gemini Robotics-ER”, the search engine giant said.
In February 2025, Apptronik raised $350m in a funding round led by B Capital and Capital Factory, with participation from Google, to further develop its humanoid robot, Apollo.
The investment will support the continued development and iteration of Apollo, expansion into new applications across industries such as logistics and manufacturing, and increased production to meet growing demand, the company said.
