Google DeepMind has revealed a new generation of the AI-powered Gemini Robotics platform. In version 2.5, Google says its robotic AI is getting tantalizingly close to completing real-world tasks as well as humans do. In fact, some Gemini robots were able to perform real jobs in Mercedes-Benz factories, and the team is preparing for the day when robots operate alongside humans with an open-source safety benchmark.
DeepMind’s physical AI platform has several computational layers between user input and robotic action. There is the vision and language model, which is Gemini Embodied Reasoning. Basically, it takes your instructions and uses live video to plan how the robot can perform the task. The other main component is vision-language-action (VLA) which actually generates robot actions, such as moving an actuator; that’s Gemini Robotics, which has moved to version 2.5.
DeepMind says that Gemini Robotics 2.5 is much better at tasks that require a high degree of dexterity. The model was evaluated on more than 200 activities, including serving food with tongs and using a pipette, as you would see in a laboratory. Google also demonstrated that Apptronik’s Apollo 2 humanoid robot with Gemini Robotics 2.5 can now make a lasso with its five-fingered hands, a notable feat, because deformable objects (like a ribbon) are notoriously difficult for robots to manipulate.
Deformable objects are difficult for robots, but Gemini is making progress.
Credit: Google
Deformable objects are difficult for robots, but Gemini is making progress.
Credit: Google
Overall, Gemini Robotics 2.5 achieves a 30 to 100 percent higher success rate compared to the latest version. It’s also 20 to 30 percent faster. But that is based on laboratory tests. Google partnered with Apptronik and Mercedes-Benz to see how capable this AI is in the face of uncertainty in the world outside of robotics labs. Gemini-powered Apollo 2 robots equipped car parts in some European factories, achieving a 96 percent success rate.
This is an impressive result compared to previous AI robotic systems, but it is probably still not enough for a company like Mercedes-Benz. A human doing that kind of repetitive task will get it right virtually every time, and DeepMind director Kanishka Rao notes that closing that last 4 percent gap will be a “steep road.”
