Google DeepMind Unveils Gemini Robotics 2 with Whole-Body Intelligence

Google DeepMind officially released Gemini Robotics 2 on August 7. As a vision-language-action (VLA) model, it provides an intelligence layer for next-generation adaptive robots, combining multimodal understanding with whole-body control. The model achieves three breakthroughs: whole-body intelligent control, advanced dexterous manipulation, and multi-robot collaboration. Official videos demonstrate interaction with a robot codenamed Apollo, highlighting natural language dialogue and embodied intelligence. This marks significant progress in whole-body coordination and collaboration for embodied AI.

Confirmed

Unconfirmed

Why it matters

Gemini Robotics 2 deeply integrates multimodal understanding with whole-body control, representing a significant advance in physical AI. Its whole-body control and multi-robot collaboration capabilities could promote the practical use of humanoid robots in complex environments, with potential impacts on industrial and service applications.

2026-08-07 ~ 2026-08-08 · 6 related posts

Full story(6 episodes)→

Primary sources

4 near-duplicate retellings: gabriberton · GoogleDeepMind · GoogleDeepMind · minchoi