AI BriefingGemini Robotics Embodied Reasoning 2 enters public preview on Google Cloud
·2026.10.27 09:00
Key point
The model achieves 91.3% accuracy in pinpointing critical video events and supports multi-robot collaboration across various embodiments.
Details
- Gemini Robotics Embodied Reasoning 2 (ER 2) is now available in public preview on the Gemini platform, allowing enterprises to deploy Google DeepMind's embodied reasoning model with Google Cloud's security and orchestration.
- The model achieves 91.3% accuracy in identifying critical video frames with a 0.96-second mean absolute frame distance, operating at 4x the execution speed and lower compute cost than larger model categories.
- ER 2 enables multi-robot collaboration and natural language interaction, allowing robots to share semantic understanding, divide labor, and be redirected by operators without technical commands.
- Developers can simulate robot behaviors in MuJoCo-Warp and NVIDIA Isaac Sim, train custom Vision-Language-Action (VLA) models using Gemini platform tools, and deploy workflows with enterprise IAM and Google Distributed Cloud.
- Early adopters include Boston Dynamics, DeepReach, Arm, and ASTAR IMRE, who are using the model for video-based inspection, data curation, robotics compute platforms, and lab automation.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
AWS Releases Guide for Building Physical AI Training Pipelines Using SageMaker and IoT Greengrass