Gemini Robotics 2.5 accelerates real-world general dexterity
Reported by Google DeepMind1 source ↓
Google DeepMind introduced Gemini Robotics 2.5, a vision-language-action model it says can learn a wider range of robot skills faster. The model is rolling out to the company’s robotics partners.
Key points
- Google DeepMind reports 30% to 100% higher task success rates and 20% to 30% faster completion than previous baselines
- An automotive kitting pilot with Apptronik and Mercedes-Benz achieved a reported 96% success rate
- The model is rolling out to Google DeepMind’s robotics partners
- Safety guidance improved safe task completion by 33% around obstacles and 43% around nearby people, according to the company
Details
Google DeepMind says it evaluated Gemini Robotics 2.5 on more than 200 tasks, including serving food with tongs, tying a bow and pipetting liquid. The company reports success rates 30% to 100% higher than previous baselines, depending on task type, and task completion 20% to 30% faster.
In a proof-of-concept with Apptronik and Mercedes-Benz, Apollo 2 humanoid robots performed automotive kitting on production lines in Kecskemét, Hungary, and Sindelfingen, Germany. Google DeepMind says the robots sorted metal brackets from bulk bins into trays in vehicle-specific order, achieving a 96% success rate.
The model works with different robot types, including Apptronik’s Apollo 2 and the Franka Duo platform, and is rolling out to Google DeepMind’s robotics partners. The company also reports that safety guidance increased safe task completion by 33% around obstacles and 43% near people.
Written from coverage as of
Images
- Google DeepMind
- Google DeepMind
- Google DeepMind
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh