Boston Dynamics’ robot dog now reads gauges and thermometers with Google’s AI

Boston Dynamics' innovative four-legged robot, Spot, has reached a new milestone in industrial automation. Thanks to Google DeepMind's latest AI model, Spot is now capable of accurately interpreting analog thermometers and pressure gauges as it navigates through factories and warehouses. Unveiled on April 14, the Gemini Robotics-ER 1.6 model is designed to enhance robotic capabilities for what is known as ‘embodied reasoning’—the ability to interact intelligently with physical environments. This high-level reasoning model enables robots to analyze complex tasks and execute them effectively. The advancements stem from a collaborative effort between Google DeepMind and Boston Dynamics, which is keen on deploying both quadruped and humanoid robots across various industrial settings, particularly within automotive manufacturing facilities owned by Hyundai Motor Group. Spot is currently being tested as a robotic inspector, tasked with performing thorough checks throughout industrial sites. These inspection duties involve intricate visual reasoning skills, allowing the robot to interpret various instruments, including liquid levels, container boundaries, and tick marks. The introduction of agentic vision, which was first featured in Google's Gemini 3.0 Flash model in January 2026, significantly enhances Spot's performance. This capability enables the robot to perform tasks with a remarkable accuracy increase—from a mere 23 percent with the previous Gemini Robotics-ER 1.5 model to an impressive 98 percent with the Gemini Robotics-ER 1.6. For context, the Gemini 3.0 Flash model achieved a 67 percent accuracy rate. Notably, even without agentic vision, the baseline Gemini Robotics-ER 1.6 can still reach an accuracy of 86 percent in instrument readings by leveraging a systematic approach to identify and analyze different visual elements. Further enhancements include an upgraded multi-view reasoning capability, allowing the robot to utilize multiple camera feeds to gain a comprehensive understanding of its surroundings.