Google DeepMind is not rushing its latest robotics breakthroughs to market. The company has placed tight restrictions on its new models, Gemini Robotics and Gemini Robotics-ER, granting access to only a small group of outside developers. The trusted testers are Agile Robots, Agility Robotics, Boston Dynamics, and Enchanted Tools. Everyone else must wait.
These models represent a significant step forward in robotics. Gemini Robotics is built on the Gemini 2.0 large language model, the same engine that powers Google’s text and image AI. This allows robots to see, understand language, and take action—a combination that marks a departure from traditional programming. Most robots can only perform tasks they are explicitly coded to do. Gemini Robotics can adapt to unfamiliar situations.
The second model, Gemini Robotics-ER, adds a layer of embodied reasoning. ER stands for embodied reasoning. This means the robot does not simply recognize an object. It understands how to interact with that object in the physical world—where to grip, how much force to apply, and what should happen next. A robot that can pick up a cup is one thing. A robot that can pick up a cup, move it to a sink, and wash it without breaking anything is another.
Google DeepMind is proceeding with caution. Restricted access allows testing in controlled environments. The four partner companies provide feedback, bugs are fixed, and safety questions are addressed before the technology spreads. The partnership with Apptronik, which builds humanoid robots, is central to this effort. Apptronik handles the hardware; DeepMind provides the software. Together, they are working to bridge the gap between a language model and a moving machine.
This is not a consumer product. It is a research platform and a proof of concept. The real effects will emerge later, in factories, warehouses, and possibly homes. A robot that understands new situations could work in a hospital, navigate a cluttered workshop, or assist in a kitchen. The potential is broad, but so is the risk of failure. A robot that misreads a situation could cause damage or injury. That is why the rollout is slow.
Boston Dynamics, one of the testers, builds some of the most advanced robots in the world. If they see value in Gemini Robotics, others may follow. The models were announced on March 12, 2025. The consequences are just beginning to unfold. Robotics companies now have a new tool. Researchers have a new target. Competitors have a new benchmark to beat.
Google DeepMind has not said when wider access will come, what the pricing will be, or which industries will get first priority. Those decisions are still being made. What is clear is that the technology works. It is based on a proven large language model, designed for real-world interaction, and being handled with care. The next few months will tell the story. If testers find problems, the rollout slows. If they find breakthroughs, pressure to open up grows. Either way, the landscape of robotics has shifted. The question is how fast the shift will spread.
{“@context”: “https://schema.org”, “@type”: “NewsArticle”, “headline”: “Google DeepMind Unveils Powerful New Robotics Models with Limited Access”, “description”: “Google DeepMind is not rushing its latest robotics breakthroughs to market. The company has placed tight restrictions on its new models, Gemini Robotics and Gem”, “datePublished”: “2026-06-21T09:30:00+00:00”, “author”: {“@type”: “Person”, “name”: “Ethan Smith”}, “publisher”: {“@type”: “Organization”, “name”: “InfoPulse Today”, “url”: “https://newsfeed.infopulsetoday.com/”}, “image”: “https://infopulsetoday.com/wp-content/uploads/sites/4/2026/09/google-deepmind-logo.png”}
