Google DeepMind's Gemini Robotics 2.5 sorts metal parts at 96% in Mercedes-Benz pilot
Reported by StartupHub AI, FourWeekMBA, The Gadgeteer and 3 more outlets6 sources ↓
Google DeepMind says Apptronik’s Apollo 2 humanoid sorted metal parts on active Mercedes-Benz production lines in Hungary and Germany, with a reported 96% success rate.
Key points
- Gemini Robotics 2.5 was evaluated on more than 200 tasks, DeepMind says
- DeepMind reports 30% to 100% higher success rates than prior baselines, depending on task, and tasks completed 20% to 30% faster
- The model is rolling out to robotics partners Agile, Apptronik and Mercedes-Benz, The Decoder reports
- The number of trials behind the 96% result was not disclosed, Tech Times reports
Sources
- StartupHub AIRead ↗Google robots hit 96% on factory kitting
- FourWeekMBARead ↗Google Robot AI Hits 96% at Mercedes; Jailbreaks Fall 36x
- The GadgeteerRead ↗It scored zero on these robot tasks. Now it aces them, and it’s working a Mercedes-Benz line
- The DecoderRead ↗Google DeepMind's Gemini Robotics 2.5 scores 96 percent sorting metal parts in a Mercedes-Benz assembly pilot
- Unite.AIRead ↗Apollo 2 Humanoids Hit 96% in Mercedes-Benz Sorting Proof of Concept
- Tech TimesRead ↗Gemini Robotics 2.5 Reports 96% Success Sorting Mercedes Car Parts but Publishes No Trial Count
Details
The pilot took place at Mercedes-Benz plants in Kecskemet, Hungary, and Sindelfingen, Germany. Apollo 2 sorted metal parts from bulk bins into containers arranged in the order needed for each vehicle. Gemini Robotics ER tracked the workflow, while Gemini Robotics 2.5 controlled the robot’s hands and arms.
DeepMind says the model was evaluated on more than 200 tasks, including handling food with tongs, tying a bow and pipetting liquid. It reports success rates 30% to 100% higher than previous baselines, depending on task type, and task completion times 20% to 30% faster. The Decoder reports that the model is rolling out to partners Agile, Apptronik and Mercedes-Benz.
Tech Times reports that DeepMind did not disclose how many trials produced the 96% figure, how it defined success or how often a person intervened. It also found no independent assessment of that result or Mercedes-Benz statement confirming it. Apptronik CEO Jeff Cardenas said the models improved after a few weeks of post-training and transferred to Mercedes production with little performance drop-off.
Written from coverage as of
Images
- The Decoder
- StartupHub AI
- Unite.AI
- Tech Times
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh