{
  "id": 19820,
  "title": "Gemini Robotics 2 brings us one step closer to physical AGI",
  "url": "https://urgent.news/2026/07/31/gemini-robotics-2-brings-us-one-step-closer-to-physical-agi",
  "topic": "ai",
  "section": "AI",
  "published": "2026-07-31T15:56:51.000Z",
  "source": {
    "name": "The New Stack",
    "slug": "the-new-stack",
    "url": "https://thenewstack.io/gemini-robotics-2/"
  },
  "original_language": "en",
  "account": "Google DeepMind unveiled Gemini Robotics 2, an intelligence layer featuring three new models, designed to imbue robots with more adaptable physical capabilities. This achievement marks a significant advancement towards large-scale Artificial General Intelligence (AGI) in the physical world. The three models – VLA, Gemini Robotics On-Device 2, and Gemini Robotics ER 2 – work together to enable robots with greater dexterity and full-body control to execute a range of multi-step tasks. While Google's Gemini Robotics 1.5 had limited control to the upper body, the new Gemini Robotics 2 now grants robots the ability to walk, crouch, stretch, and manipulate various objects, allowing them to perform complex tasks like fetching items from high shelves. In addition to physical agility, the VLA model enables robots to process vision, language, and action inputs to coordinate limbs and execute intricate tasks with the finesse of a human hand. The ER model, on the other hand, acts as the robot's \"high-level brain,\" allowing it to communicate with humans, process instructions, and coordinate actions while continuously adapting to changes in the environment. This continuous perception loop mirrors the way human brains process spatial information, enabling robots to handle unexpected obstacles, battery drainage, or changes in the surrounding environment. Google also introduced ASIMOV-Agentic, a new safety benchmark to ensure robots can refuse unsafe tasks and seek human intervention when needed. Gemini Robotics 2 is currently available through Google AI Studio and private preview on Gemini Enterprise Agent Platform, with VLA models accessible to early-access partners.",
  "summary": "This week, Google DeepMind revealed Gemini Robotics 2, an intelligence layer comprising three new models to power more adaptable physical The post Gemini Robotics 2 brings us one step closer to physical AGI appeared first on The New Stack .",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 3,
    "also_reported_by": [
      {
        "outlet": "Google DeepMind",
        "title": "Gemini Robotics 2 brings whole body intelligence to robots",
        "url": "https://urgent.news/2026/07/28/gemini-robotics-2-brings-whole-body-intelligence-to-robots",
        "published": "2026-07-28T13:21:37.000Z"
      },
      {
        "outlet": "Wired Business",
        "title": "Gemini Robotics 2 Brings Google's AI Into the Physical World",
        "url": "https://urgent.news/2026/07/30/gemini-robotics-2-brings-googles-ai-into-the-physical-world",
        "published": "2026-07-30T15:04:22.000Z"
      }
    ]
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}