DeepMind Delivers Gemini Robotics 2.5 to Robotics Partners, Apollo 2 Hits 96% Sorting Success

Google DeepMind has released Gemini Robotics 2.5, described as its most capable vision-language-action system to date, and is rolling it out to robotics partners. It reports success rates 30% to 100% higher than earlier baselines and 20% to 30% faster task completion across more than 200 tasks, plus a proof-of-concept with Apptronik and Mercedes-Benz doing automotive kitting on production lines.

Google DeepMind released Gemini Robotics 2.5 on November 10, describing it as its most capable vision-language-action model to date. Google says it is now available to its robotics partners, with 96% sorting success in a pilot at a Mercedes factory.

Google DeepMind released Gemini Robotics 2.5 on November 10 local time, describing it as its most capable vision-language-action robotics model yet and opening it to robotics partners. According to the company's blog, the model improves success rates by up to 100% and speed by up to 30% on some tasks, and reaches 96% success on parts sorting at Mercedes-Benz lines in Hungary and Germany.

Google DeepMind has officially released Gemini Robotics 2.5, its most capable model yet for translating visual perception into physical action. In tests on Mercedes-Benz production lines, Apptronik Apollo 2 humanoid robots powered by the system achieved a 96% success rate in sorting car parts.

Google released Gemini Robotics 2.5, a vision-language-action model it says improves dexterity, reliability, and safety, tested across more than 200 tasks with success rates 30%–100% higher and 20%–30% faster completion than the prior baseline. In a Mercedes production-line proof of concept, it drove Apptronik's Apollo 2 humanoid to pick and pack parts with a 96% success rate. The model is opening to Google's robotics partners.

Google DeepMind introduced Gemini Robotics 2.5 on November 10, calling it its most capable vision-language-action model for robots yet, able to pick up dexterous tasks faster and operate in less predictable real-world settings. Across more than 200 evaluated tasks, DeepMind reports success rates 30% to 100% higher than previous baselines and 20% to 30% faster evaluation, with tasks like removing a disk drive from its bay and placing a lime on a top shelf going from 0% to 100%.

Google DeepMind introduced Gemini Robotics 2.5, a new robot control model, and tested it with Apptronik on active Mercedes-Benz production lines in Hungary and Germany, The Decoder reports. The humanoid robot Apollo 2 sorted metal parts there with a reported 96% success rate.

Google announced Gemini Robotics 2.5, a new model for physical AI that was built with support from Gemini 4 Argon, according to TestingCatalog. In a proof of concept with Apptronik and Mercedes-Benz, the Apollo 2 robot used it to pick and organize car parts for assembly with 96% success.

DeepMind has announced Gemini Robotics 2.5, a robotics model the company says is up to twice as successful at some tasks, Ars Technica reports. The announcement also includes an open safety benchmark called ASIMOV-jailbreak.

Google DeepMind released Gemini Robotics 2.5, calling it its most capable vision-language-action model to date. The company says the update improves robot dexterity, reliability and safety in real-world settings. Vision-language-action models translate visual input and instructions directly into robot actions.