Google Debuts Gemini Robotics-ER 1.5, Bringing Embodied Reasoning to Real-World Robots

September 25, 2025

Details

  • Google AI Developers revealed Gemini Robotics-ER 1.5 on September 25, 2025, now available in preview through Google AI Studio.
  • This model is a purpose-tuned variant of Gemini 1.5, expanding its existing language and vision abilities with spatial reasoning and motor control specifically optimized for robotics.
  • Key enhancements include advanced autonomous behaviors, a configurable “thinking budget” to balance latency against depth of planning, and stronger safety filters to prevent hazardous robotic commands.
  • Developers can access it immediately with AI Studio workspaces, REST API endpoints, comprehensive documentation, and a public Colab notebook featuring ROS 2 examples.
  • The model leverages Google's latest TPU-Edge accelerators for efficient on-device use, supporting mobile manipulation and warehouse automation pilots.
  • This milestone follows recent Gemini 1.5 multimodal updates, signaling Google’s first concerted move to commercialize embodied reasoning in robotics.

Impact

Google's launch intensifies the rivalry with OpenAI's GPT-4R and NVIDIA’s Isaac agents, heightening the race to claim robotics developers. The robust preview access and open tools could accelerate robot use in industries like logistics and eldercare. By focusing on safety, compliance, and hardware integration, Google positions itself as a leading force as the sector pushes toward safer, more autonomous robots.

Read more