Google DeepMind Launches Gemini Robotics 2 for Whole-Body Control and Dexterous Manipulation
量子位 · wechat · 2026-07-31
Google DeepMind has unveiled its next-generation robot brain, Gemini Robotics 2, alongside an advanced reasoning model (ER2) and an on-device model (On-Device 2), completing a full stack for embodied AI.
- Gemini Robotics 2 (VLA Model): The core breakthrough is unified whole-body control, moving beyond desktop tasks. It enables full-size humanoids to walk, bend, and perform dexterous manipulations (like unscrewing lightbulbs or zipping bags) using multi-fingered hands.
- Gemini Robotics ER2 (Advanced Brain): Handles instruction understanding, decomposing long-horizon multi-step tasks, and tracking progress. It features error re-planning and multi-robot collaboration capabilities.
- Gemini Robotics On-Device 2: Runs locally without cloud access and can adapt to new dual-arm robot embodiments in hours using fewer than 200 examples.
Demos highlighted impressive generalization across various hardware platforms, including the Apptronik Apollo 2 humanoid.
Related event: Google DeepMind Unveils Gemini Robotics 2(48 posts)→
More from Embodied
- DeepMind Launches Gemini Robotics 2: Enabling Whole-Body Intelligence and Multi-Robot Collaboration — keerthanpg · 2026-07-31
- Unitree's Aug 10 IPO: How the Market Prices China's Next Productivity Leap — yongqianme · 2026-07-31
- FCC Bans Foreign Humanoid Robots; US Maker Offers Sub-$2k Hardware — scott_e_reed · 2026-07-31
- RKO-LIO: Open-Sourcing Robust LiDAR-Inertial Odometry Without Sensor-Specific Modelling — tom_doerr · 2026-07-31
- US Rules Foreign-Made Robots Pose National Security Risk, Mandating Domestic Production — ZeroStateReflex · 2026-07-31
- AgenticROS Debuts Cube Pick-and-Place Skill for LeRobot Robotic Arm — chrismatthieu · 2026-07-31