Boston Dynamics and Google DeepMind have partnered to enhance their robot Spot by integrating Gemini Robotics, allowing it to understand its environment and respond to natural language commands. This innovation emerged from a 2025 hackathon aimed at leveraging Large Language Models (LLMs) and Visual Foundation Models (VFMs) to enable more complex tasks, moving beyond traditional programming methods. The collaboration aims to simplify how developers interact with Spot, making it easier to use conversational prompts to issue commands, thereby streamlining the process of robot operation and expanding the potential applications of such technology in various fields.
Spot: Spot is Boston Dynamics’ quadruped robot designed for navigation, sensing, and manipulation in challenging settings like factories and homes. It features an SDK and API for custom applications in inspection and research. In the partnership demo, Spot used Gemini Robotics to autonomously tidy rooms by interpreting high-level instructions and sequencing API calls.
Boston Dynamics: Boston Dynamics develops advanced mobile robots including the quadruped Spot for industrial inspection, research, and data analysis in rugged environments. They recently formalized a partnership with Google DeepMind to integrate AI models into their robots. This collaboration enables Spot to execute complex tasks through natural language instructions via Gemini Robotics.
Gemini Robotics: Gemini Robotics is Google’s suite of visual-language AI models tailored for robotics applications, emphasizing embodied reasoning to process environments and generate actions. In the news, it bridges natural language prompts to Spot’s controls, allowing conversational task execution like object identification and manipulation. The models power demos and tools like AIVI-Learning for enhanced robot intelligence.
Gemini Robotics ER 1.5: Gemini Robotics ER 1.5 is a Google visual-language model focused on embodied reasoning for robotics, enabling environmental contextualization and action planning without state-machine code. It powered the recent Spot demo originating from a hackathon, where it processed images and directed tasks like picking up objects. The model relies on detailed tool prompts to interface with robot APIs effectively.
Gemini Robotics ER 1.6: Gemini Robotics ER 1.6 represents an evolved version of Google’s robotics AI model with improved visual intelligence for inspection tools. It has been rolled out in Boston Dynamics’ AIVI-Learning software for Spot and Orbit robots. This deployment adds automatic enhancements, bringing deeper contextual understanding to industrial applications.
`json
{
“Partnership”: “Boston Dynamics and Google DeepMind collaborated to integrate Gemini Robotics models into robots like Spot to facilitate natural language control.”,
“Hackathon Origin”: “The Spot demonstration originated from a Boston Dynamics hackathon leveraging large language and visual foundation models for autonomous robot capabilities.”,
“Developer Benefit”: “AI models like Gemini Robotics support rapid experimentation with Spot by converting conversational commands into executable actions via its API.”,
“Product Enhancement”: “The AIVI-Learning tool now utilizes the Gemini Robotics ER 1.6 model for enhanced visual inspection functionalities on Spot and Orbit.”
}
`