News

Tag: robots

  • Gemini Robotics 2.5 Hits 96% on a Real Mercedes Line

    Gemini Robotics 2.5 Hits 96% on a Real Mercedes Line

    Google DeepMind has unveiled Gemini Robotics 2.5, its most capable model yet for controlling robots, and the headline test did not take place in a laboratory. Apptronik’s Apollo 2 humanoid robots, running the new model, worked on live Mercedes-Benz production lines in Hungary and Germany, picking metal car parts from bulk bins and packing them into sequencing containers in the order each vehicle needed them. DeepMind reported a 96% success rate for the work.

    The pilot ran at Mercedes-Benz plants in Kecskemét, Hungary, and Sindelfingen, Germany. A second DeepMind model, Gemini Robotics ER, kept track of the workflow while Gemini Robotics 2.5 controlled the robot’s hands and arms, correcting for gaps as parts were moved. The job is known in factories as kitting: gathering the right parts, in the right order, before assembly begins.

    What the new model changes

    DeepMind describes Gemini Robotics 2.5 as a vision-language-action model, which means it connects what a robot sees and what it is told to the movements it makes. The company says it tested the model on more than 200 tasks, with success rates 30% to 100% higher than earlier baselines depending on the task, and tasks finished 20% to 30% faster. It also says the model can control whole humanoid bodies, including stepping, squatting and bending to keep balance.

    The 96% figure deserves a note of care. DeepMind has not published how many trials the number covers, how it defines a successful pick, or how often a person stepped in, and no independent check of the figure has been published. A proof of concept on one job, however real the factory floor, is not the same as a robot working a full shift unsupervised, and Mercedes-Benz has not said it plans to buy fleets of the robots.

    The model is rolling out only to DeepMind’s robotics partners, including Agile Robots, Apptronik and Boston Dynamics, along with more than 100 trusted testers. There is no public version, no price and no date for wider availability. Apptronik said the models improved sharply within a few weeks of post-training and carried over to the Mercedes-Benz lines with very little drop-off.

    Alongside the model, DeepMind released an open safety benchmark called ASIMOV-Jailbreak, which tests whether a robot can be talked into unsafe actions. For now, the Mercedes-Benz pilot is best read as a sign of how far humanoid robots have moved from demonstrations towards real factory work, and how much testing still lies between a promising trial and a production line that depends on a robot every day.

  • Meet Robia and Jojo, Karachi’s Robot Waitresses

    Meet Robia and Jojo, Karachi’s Robot Waitresses

    In a busy Karachi dining room, the newest member of the wait staff glides between booths on wheels, pauses at a table, and speaks. “Welcome to Sicilian. My name is Robia,” she says in an electronic voice, her red LED eyes blinking as diners look up from their menus. Beside her works Jojo, the second of two humanoid robot waitresses now serving guests at The Sicilian restaurant.

    The pair have quickly become the restaurant’s star attraction. Shaped with spherical white heads and dark visors that resemble a motorcycle helmet, the robots wear black shawls around their necks in the style of a dupatta. They move from table to table carrying plates — pizza among the dishes — and when they arrive, customers lift the food off their serving trays themselves or wait for a human member of staff to step in.

    Behind the spectacle is a homegrown story. The robots were built by Pakistani engineer Syed Osama Aziz, who also founded the restaurant, through his robotics company Cobot. Most of the parts were manufactured locally, with the circuit boards designed and programmed in his own laboratory, although some electronic components were imported from abroad. His workshop is based in Multan, in the province of Punjab.

    A restaurant that doubles as a showcase

    For Aziz, the restaurant is as much a demonstration of what Pakistani engineering can do as it is a place to eat. The machines are built around locally made components, a point of pride for a project that shows off domestic manufacturing rather than relying on imported finished robots. Guests who come for the food often end up filming the servers instead, and the robots have turned an ordinary dinner into something people talk about long after they leave.

    Yet the robots are far from finished. Aziz plans to add camera-based processing and image-processing features so that a robot can recognise a customer and remember what they ordered last time. The idea is simple: a returning guest would be greeted not only by name, but with suggestions shaped by their previous visit.

    That ambition points to a wider trend. Around the world, restaurants have experimented with machines to carry food, take orders and cut waiting times, but most of those systems arrive ready-made from large overseas manufacturers. Here, the machines were designed and assembled close to home, by the same person who opened the restaurant they work in.

    For now, Robia and Jojo keep to their rounds: wheeling through the dining room, stopping at each booth, and letting diners help themselves from the tray. The human staff remain close by, stepping in whenever a guest needs a hand. The robots may not yet remember your last order — but if their creator’s plans work out, they soon will.