Google DeepMind has launched an on-device version of its Gemini Robotics AI model, a vision-language-action (VLA) model designed to operate directly on robots without requiring an internet connection. This efficient new model, which performs nearly as well as its flagship hybrid counterpart, enables robots to generalize tasks, understand commands, and execute fine motor skills in diverse environments, including those with poor connectivity or high security needs. The accompanying SDK release signals Google's push for broader developer engagement and potential commercialization, marking a significant step towards more autonomous and adaptable robotic solutions.
Google DeepMind has released an on-device version of its Gemini Robotics AI model, a strategic move that significantly expands the operational scope of its robotics platform. By enabling the vision-language-action (VLA) model to function without an internet connection, Google is targeting applications in environments with poor connectivity or stringent security requirements. The on-device model's performance is reported to be nearly on par with the flagship hybrid model, and its adaptability is notable, requiring as few as 50 to 100 demonstrations to learn new tasks and having already been deployed on various third-party robots like Apptronik's Apollo. The concurrent release of a software development kit (SDK)—a first for a DeepMind VLA—is a critical step toward commercialization. This signals a strategy to build a developer ecosystem, fostering broader adoption and innovation beyond Google's internal projects and positioning the company to capture value in the nascent but high-potential market for autonomous robotic systems.
AI-powered research, real-time alerts, and portfolio analytics for institutional investors.
Request a DemoOverall Sentiment
strongly positive
Sentiment Score
0.65
Ticker Sentiment
