{"slug": "google-deepmind-releases-gemini-robotics-2-5-claims-96-success-on-mercedes-benz", "title": "Google DeepMind Releases Gemini Robotics 2.5, Claims 96% Success on Mercedes-Benz Factory Pilot", "summary": "Google DeepMind announced Gemini Robotics 2.5 on November 10, a vision-language-action model rolling out to robotics partners Agile, Apptronik, and Mercedes-Benz rather than the public, after a proof-of-concept pilot in which Apptronik's Apollo 2 humanoid robots performed automotive kitting at Mercedes-Benz plants in Kecskemét, Hungary, and Sindelfingen, Germany, with a reported 96% success rate. DeepMind reports the model beat unnamed previous baselines by 30% to 100% on over 200 distinct tasks and completed them 20% to 30% faster, while inference-time reasoning guidance raised safe task completion by 33% around obstacles and 43% around nearby people. DeepMind director Kanishka Rao told Ars Technica that closing the remaining 4% gap would be a \"steep road,\" and the release includes no public API, pricing, technical paper, or independent evaluation.", "body_md": "# Google DeepMind Releases Gemini Robotics 2.5, Claims 96% Success on Mercedes-Benz Factory Pilot\n\nGoogle DeepMind announced Gemini Robotics 2.5 on November 10, calling it its most capable vision-language-action (VLA) model to date. The model is rolling out to robotics partners — not the public — after a proof-of-concept run on Mercedes-Benz production lines where Apptronik's Apollo 2 humanoid robots picked and packed automotive parts with a reported 96% success rate.\n\n## What shipped\n\nGemini Robotics 2.5 is a VLA model: it takes visual input and language instructions and outputs robot actions. DeepMind says it improves on three axes — dexterity, reliability, and safety — and that Gemini 4 Argon, a separate Gemini model, accelerated its development by generating simulated practice runs from real camera scenes.\n\nThe announcement names three partners: Agile, Apptronik, and Mercedes-Benz. No price, public API, or technical paper was released. DeepMind's blog post states the model \"is rolling out to our robotics partners.\"\n\n## The Mercedes-Benz pilot\n\nThe headline reliability figure comes from a proof of concept at Mercedes-Benz plants in Kecskemét, Hungary, and Sindelfingen, Germany. Apollo 2 robots performed automotive kitting: reaching into bulk bins, grasping metal brackets, and placing them into sequencing trays in vehicle-build order.\n\nTwo models divided the work. Gemini Robotics ER (embodied reasoning) tracked the multi-step workflow. Gemini Robotics 2.5 controlled the robot's hands and arms. DeepMind reports a 96% success rate on this task.\n\nDeepMind director Kanishka Rao told Ars Technica that closing the remaining 4% gap would be a \"steep road.\" Robotics head Carolina Parada described a tension between making a robot excellent at one application and preserving its ability to handle many different jobs, saying that pushing for generalization may mean not reaching 99.9% reliability in every application [\\[6\\]](https://superpowerdaily.com/posts/google-deepmind-releases-robot-ai-citing-96-success-in-mercedes-factory-pilot).\n\nThe post does not say how many picks the 96% covers, nor how long the pilot ran. It is labeled a proof of concept, not a production deployment.\n\n## Lab results: 200 tasks, unnamed baselines\n\nDeepMind evaluated the model on \"over 200 distinct tasks\" including serving food with tongs, tying a bow on a gift box, and pipetting liquid. It reports a 30% to 100% higher success rate than \"previous baselines\" depending on task type, and 20% to 30% faster completion.\n\nThe baselines are not named, and their absolute scores are not given. The post says the model works on Apptronik's Apollo 2 (with a 22-degree-of-freedom SharpaWave hand) and the Franka Duo dual-arm platform with two-finger grippers.\n\nArs Technica reports the model still runs on cloud servers, meaning the reported speed gains do not establish low-latency, on-robot operation.\n\n## Safety: steerability and a new jailbreak benchmark\n\nDeepMind says Gemini Robotics 2.5 uses inference-time reasoning guidance to steer around hazards in real time. In its evaluations, this boosted safe task completion by 33% around obstacles and 43% around nearby people.\n\nThe company is also releasing ASIMOV-Jailbreak, an open-source benchmark of hundreds of prompts that disguise unsafe requests across dozens of scenes. DeepMind reports 99.5% jailbreak resistance for Gemini 4 Argon on this benchmark, compared to 82% for Gemini Robotics ER 2.\n\nThe 99.5% figure belongs to Argon, not to Gemini Robotics 2.5 itself. The post gives no ASIMOV-Jailbreak score for the VLA model being released.\n\n## What is not in the release\n\n- No public access, pricing, or API documentation.\n- No named baselines for the 30–100% improvement claim.\n- No sample size for the 96% factory figure.\n- No technical paper for Gemini Robotics 2.5 (a Gemini 4 safety report is linked).\n- Every number comes from DeepMind or its partners; no independent evaluation is cited.\n\n## What to watch\n\n- Whether ASIMOV-Jailbreak scores from other labs match DeepMind's.\n- Whether the Mercedes pilot extends to sustained, shift-long operation with recovery from grasp failures.\n- Whether partners beyond the named three receive access, and on what terms.\n- Whether on-robot compute replaces cloud inference for latency-sensitive tasks.\n\nDeepMind calls Gemini Robotics 2.5 \"an important milestone on the path toward solving physical AGI.\" The factory pilot is a concrete step. The gap between a 96% proof of concept and a robot that stays on the line remains, as DeepMind's own researchers acknowledge, a steep road.\n\n## Sources\n\n1. [Gemini Robotics 2.5: Accelerating Physical Intelligence](https://deepmind.google/blog/gemini-robotics-2-5-accelerates-real-world-physical-intelligence) (deepmind.google)\n2. [Gemini Robotics 2.5: 200 Tasks, 96% on a Mercedes Line](https://cellcog.ai/blog/gemini-robotics-2-5) (cellcog.ai)\n3. [Gemini Robotics 2.5: 96% Kitting, New Safety Bench](https://explainx.ai/blog/gemini-robotics-2-5-mercedes-kitting-asimov-jailbreak-2026) (explainx.ai)\n4. [Google brings Gemini Robotics 2.5 to a new factory test](https://www.blackscarab.ai/news/google-gemini-robotics-2-5-factory-kitting) (blackscarab.ai)\n5. [Gemini Robotics 2.5 launches with enhanced everyday skills](https://cryptobriefing.com/gemini-robotics-2-5-launch-everyday-skills) (cryptobriefing.com)\n6. [Google DeepMind releases robot AI, citing 96% success in Mercedes factory pilot](https://superpowerdaily.com/posts/google-deepmind-releases-robot-ai-citing-96-success-in-mercedes-factory-pilot) (superpowerdaily.com)\n7. [DeepMind unveils Gemini Robotics 2.5 and open safety benchmark ASIMOV-Jailbreak | Межа](https://mezha.net/eng/news/mezha-76165dd7fb3fabecc2835bf4_deepmind_unveils_gemini) (mezha.net)", "url": "https://wpnews.pro/news/google-deepmind-releases-gemini-robotics-2-5-claims-96-success-on-mercedes-benz", "canonical_source": "https://epyhia.com/google-deepmind-releases-gemini-robotics-2-5-claims-96-success-on", "published_at": "2026-11-10 18:17:40+00:00", "updated_at": "2026-11-10 18:20:47.919293+00:00", "lang": "en", "topics": ["robotics", "artificial-intelligence", "ai-safety", "large-language-models"], "entities": ["Google DeepMind", "Gemini Robotics 2.5", "Mercedes-Benz", "Apptronik", "Apollo 2", "Kanishka Rao", "Carolina Parada", "ASIMOV-Jailbreak"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/google-deepmind-releases-gemini-robotics-2-5-claims-96-success-on-mercedes-benz", "markdown": "https://wpnews.pro/news/google-deepmind-releases-gemini-robotics-2-5-claims-96-success-on-mercedes-benz.md", "text": "https://wpnews.pro/news/google-deepmind-releases-gemini-robotics-2-5-claims-96-success-on-mercedes-benz.txt", "jsonld": "https://wpnews.pro/news/google-deepmind-releases-gemini-robotics-2-5-claims-96-success-on-mercedes-benz.jsonld"}}