Gemini Robotics 2.5: Google DeepMind's vision-language-action model runs humanoids on Mercedes-Benz lines

Google DeepMind introduced Gemini Robotics 2.5 on November 10, 2026, calling it its most capable vision-language-action (VLA) model to date. The model turns what a robot sees and is told into arm and hand movements, and DeepMind says it widens the range of skills a robot can learn while speeding up how fast it learns them. It builds on the Gemini Robotics line that gained whole-body control earlier this year, and DeepMind credits Gemini 4 Argon with accelerating the research. Gemini Robotics 2.5 is rolling out to DeepMind's robotics partners; there is no public API or price.

DeepMind evaluated the model across more than 200 distinct tasks, including serving food with tongs, tying a bow on a gift box and pipetting a liquid. Compared with its previous baselines, it reports success rates 30% to 100% higher depending on the task type, with tasks finishing 20% to 30% faster; all figures are DeepMind's own, with no independent evaluation yet. The model runs on different robot bodies, including Apptronik's Apollo 2 humanoid with a five-finger hand and the two-armed Franka Duo. Inference-time reasoning guidance raised safe task completion around obstacles by 33% and around people by 43%, and DeepMind released ASIMOV-Jailbreak, an open-source benchmark of disguised unsafe requests, on which Gemini 4 Argon scores 99.5% jailbreak resistance against 82% for Gemini Robotics ER 2.

The headline test ran outside the lab: Apollo 2 humanoids running Gemini Robotics 2.5 sorted automotive parts on active Mercedes-Benz production lines in Kecskemét, Hungary, and Sindelfingen, Germany. The robots performed automotive kitting — picking metal brackets from bulk bins and placing them into sequencing trays — with a 96% success rate, using the Gemini Robotics ER model to track the workflow while the 2.5 actions model controlled the hands and arms.