Google’s latest vision-language-action model for robotics is now being deployed to its partners following internal tests that saw task success rates jump by 30% to 100% over previous baselines. Gemini Robotics 2.5 was evaluated across more than 200 distinct manual tasks, including tying gift bows and pipetting liquids, and recorded a 20% to 30% increase in the speed of task completion.
The model was developed over four months using Gemini 4 Argon as a simulation engine to generate virtual practice runs and refine the training process. In a real-world pilot at Mercedes-Benz plants in Germany and Hungary, the software powered Apptronik’s Apollo 2 humanoid robots to sort metal brackets on production lines with 96% accuracy.
