Google introduces Gemini Robotics 2.5
On November 10, 2026, Google DeepMind introduced Gemini Robotics 2.5, a vision-language-action model rolling out to robotics partners.
- A vision-language-action model for controlling robots across different tasks.
- Google reports more than 200 evaluated tasks and an industrial pilot that sorted parts with 96% success.
- Source and date: Google DeepMind, November 10, 2026; deployment and results are still described as a pilot.
Context and analysis by Noosigma AI
Editorial summary and context based on the cited source. Claims and results attributed to the organization are reports from that source; they do not constitute independent validation by Noosigma AI.
The company says the system broadens the set of physical tasks robots can learn and improves delicate, multi-step actions. Its published evaluation covered more than 200 distinct tasks and included tests on different robot platforms.
Why it matters
In a pilot with Apptronik and Mercedes-Benz, an Apollo 2 robot sorted automotive parts into trays on a production line, achieving a 96% success rate according to Google. The article describes this as a proof of concept at specific sites, not as general production performance.
The release includes physical safety controls and an evaluation of robotics-focused jailbreak attempts. The stated rollout is to partners; the figures and comparisons are evaluations published by Google DeepMind. Source and date: Google DeepMind, November 10, 2026.