Google DeepMind has officially announced Gemini Robotics 2, a suite of artificial intelligence models designed to deliver whole-body motor control, local execution, and multi-robot orchestration. Introduced by DeepMind's Carolina Parada on July 30, 2026, the launch follows the April 14, 2026 announcement of Gemini Robotics-ER 1.6 by Laura Graesser and Peng Xu.
The new suite aims to transform how autonomous systems interact with physical environments. By unifying perception, spatial planning, and physical movement, the models allow hardware platforms to execute complex tasks with fine dexterity and reasoning capabilities.
Four Models Power Google DeepMind's Embodied Intelligence Architecture
The Gemini Robotics lineup is split into dedicated vision-language-action (VLA) and embodied reasoning (ER) architectures. Each model targets a distinct operational layer in robotic systems, ranging from high-level multi-agent orchestration to low-latency local execution.
| Model Name | Model Type | Primary Function | Announcement Date |
|---|---|---|---|
| Gemini Robotics 2 | Vision-Language-Action (VLA) | Whole-body motor control and fine physical dexterity | July 30, 2026 |
| Gemini Robotics ER 2 | Embodied Reasoning (ER) | Task planning, video understanding, and multi-robot collaboration | July 30, 2026 |
| Gemini Robotics On-Device 2 | Local VLA | Efficient, low-latency local model processing | July 30, 2026 |
| Gemini Robotics-ER 1.6 | Embodied Reasoning (ER) | Spatial reasoning, instrument reading, and multi-view understanding | April 14, 2026 |
Whole-Body Intelligence Enables Fine Dexterity and Multi-Robot Collaboration
Google DeepMind designed Gemini Robotics 2 as a VLA model capable of controlling hardware systems from feet to fingertips. Rather than isolating control to individual arms or manipulators, the model generates coordinated actions across an entire physical frame.
For complex operational workflows, Gemini Robotics ER 2 introduces advanced reasoning capabilities designed for physical environment understanding. Carolina Parada noted that Gemini Robotics ER 2 represents a step change in powering robots with video understanding, task orchestration, and multi-robot collaboration.
This release builds directly on the foundation laid by Gemini Robotics-ER 1.6 earlier in the year. That model focused on enhancing multi-view understanding, spatial precision, and instrument reading capabilities for real-world robotics tasks.
Discussion
0 commentsNo comments yet. Be the first to share your take.