FULL STORY

Google Launches Gemini Robotics 2

Following early leaks, Google DeepMind officially launched Gemini Robotics 2, introducing its new vision-language-action and embodied reasoning models.

2026-07-30 ~ 2026-07-31 · 2 episodes · 22 posts

Episode 1 · Google DeepMind Teases Gemini Robotics 2.0 (2026-07-30, 2 posts)

Google DeepMind is gearing up for a major 2.0 update to its Gemini Robotics model. Recent leaks hint at new progress and demonstrations, building anticipation following its initial April release.

Episode 2 · Google DeepMind Unveils Gemini Robotics 2 (2026-07-30, 20 posts)

Google DeepMind officially releases Gemini Robotics 2, advancing embodied AI from tabletop tasks to full-body control. The new model equips humanoid robots with advanced dexterous manipulation and whole-body motor coordination, while introducing multi-robot collaboration to tackle complex workflows.

已确认

  • 核心发布：Google DeepMind unveiled Gemini Robotics 2, a new Vision-Language-Action (VLA) model designed to provide an intelligence layer for highly adaptive robots, breaking the limitations of traditional pre-programming or teleoperation.
  • 全身控制与灵巧操作：The new model unlocks intelligent whole-body motor control, enabling robots to stretch, bend, and balance in tight, cluttered spaces. It also features advanced dexterous manipulation across various end-effectors (such as robotic hands or grippers), allowing for precise actions like tying knots and screwing in lightbulbs.
  • 多机协作：A multi-robot team collaboration feature is introduced, allowing different types of robots to communicate and cooperate to solve complex tasks that a single unit cannot accomplish independently.
  • 配套模型：The embodied reasoning model Gemini Robotics-ER was launched simultaneously, forming a model matrix that drives breakthroughs in embodied intelligence.

为什么重要

  • This release marks the transition of robotic intelligence from isolated tabletop tasks to full-body intelligence adapted for real-world human environments. By enhancing comprehensive perception, motor coordination, and multi-robot collaboration, the new system is expected to play a more practical assistive role in complex and unknown environments like homes and workplaces.