Google DeepMind announced Gemini Robotics 2.5 on November 10, 2026. It describes the model as its most capable vision-language-action system so far, and says it is rolling out to its robotics partners.
What the model does
DeepMind evaluated the system across over 200 different tasks, among them using tongs to plate food, knotting a ribbon on a present box, and dispensing a fluid with a pipette. Against earlier baselines it reports a success rate 30% to 100% higher, depending on the task type, and tasks finished 20% to 30% faster. The model works with several robot types, including Apptronik's Apollo 2 humanoid and the Franka Duo platform with two-fingered grippers.
The factory pilot
The main real-world evidence is a proof-of-concept with Apptronik and Mercedes-Benz on active production lines in Kecskemét, Hungary, and Sindelfingen, Germany. Apollo 2 did automotive kitting: it took metal brackets out of bulk bins and placed them into sequencing containers in the order each vehicle needs. Gemini Robotics ER tracked the workflow, while Gemini Robotics 2.5 controlled the hands and arms. The reported success rate was 96%. Apptronik's CEO said the models improved sharply after a few weeks of post-training and carried over to the Mercedes-Benz lines with little loss. These figures come from the vendors and are not independently verified.
Safety claims
The model uses reasoning at inference time to steer around physical hazards. DeepMind says this raised safe task completion by 33% around obstacles and by 43% around nearby people. It also released ASIMOV-Jailbreak, an open-source benchmark of hundreds of prompts that hide unsafe requests across dozens of scenes. According to DeepMind, Gemini 4 Argon resists 99.5% of them, against 82% for the earlier Gemini Robotics ER 2.
Argon as a training tool
DeepMind says Gemini 4 Argon sped up development by acting as a simulation engine. It can produce thousands of varied practice runs from a single real camera scene and build a digital twin of it. It can also suggest why a task failed from video footage and test fixes in simulation and on real robots.
What is not yet known
The announcement names no pricing, no general availability date and no list of partners beyond Apptronik, Agile and Mercedes-Benz in the acknowledgements. The 96% figure covers one kitting task, so it says little about other jobs.
Top comments (0)