The first model, **Gemini Robotics**, is a vision-language-action system designed to enable robots to perform physical actions as part of their output. The second, **Gemini Robotics-ER**, enhances a robot’s spatial awareness and allows developers to integrate their own programs using the reasoning capabilities of Gemini 2.0.
Google stated that these models are adaptable to robots of all types, including humanoids and those used in factories and warehouses. The company tested the Gemini Robotics model on its bi-arm robotics platform, ALOHA 2, but emphasized its potential for more complex applications, such as Apptronik’s Apollo robot.
The launch follows recent developments in the robotics sector, including robotics startup Figure AI’s decision to end its collaboration with OpenAI last month after achieving an internal breakthrough in AI for robotics. Google’s move also aligns with its broader strategy to support the robotics industry, as evidenced by its participation in Apptronik’s recent $350 million funding round, led by B Capital and Capital Factory, to scale production of AI-powered humanoid robots.
Google’s foray into robotics is not new. The company acquired robotics pioneer Boston Dynamics in 2013 but sold it to SoftBank Group Corp four years later. Boston Dynamics is renowned for its advanced humanoid and dog-like robots.
By offering robotics-focused AI models, Google aims to help startups reduce development costs and accelerate time-to-market for their products. This initiative underscores the growing synergy between AI and robotics, as both fields continue to evolve and drive innovation.