AI tools ·
Google's Gemini Robotics 2.5 hit 96% packing car parts on a live Mercedes line
Google DeepMind's newest robot brain is already working a real factory floor. Gemini Robotics 2.5, announced November 10, is the lab's most capable vision-language-action model — the kind that turns what a robot sees and is told into movement. In a proof of concept on active Mercedes-Benz production lines, Apptronik's Apollo 2 humanoid robots picked and packed automotive parts with a 96% success rate.
DeepMind tested the model across more than 200 distinct tasks — serving food with tongs, tying a bow on a gift box, pipetting liquid — and reports success rates 30% to 100% higher than earlier baselines, with tasks completed 20% to 30% faster. It runs on Apollo 2's five-finger hand and on the Franka Duo two-arm platform. The Mercedes pilot ran in Kecskemét, Hungary, and Sindelfingen, Germany, where the robots sorted parts from bulk bins into trays in the exact order each vehicle on the line needed.
The catch for everyone else: there is no public access yet. DeepMind says Robotics 2.5 is rolling out to its robotics partners — Agile Robots, Apptronik, and Boston Dynamics — with no price or general-availability date. The safety work is public, though: DeepMind released ASIMOV-Jailbreak, an open benchmark that tests whether robots can be tricked into unsafe actions, and says the model steers around physical hazards in real time.
Why it matters
A 96% success rate on a live production line — not a lab demo — is the number that moves humanoid robots from research projects toward real jobs. The monotonous, physically taxing work is exactly what factories want to automate first, and DeepMind's partner-only rollout means the first deployments will be inside companies like Mercedes, not on your doorstep. The open safety benchmark matters too: as robots get more capable, testing whether they can be jailbroken becomes as important as testing what they can do.
Sources
- Google DeepMind — Gemini Robotics 2.5: Accelerating Physical Intelligence
- TechTimes — Gemini Robotics 2.5 Reports 96% Success Sorting Mercedes Car Parts
- Unite.ai — Apollo 2 Humanoids Hit 96% in Mercedes-Benz Sorting Proof of Concept
Spot an error? We correct quickly and note it. Contact us via the contact page.