Our Gemini-based multimodal model gives advanced world understanding to robots.
It acts as a high-level brain to break down complex tasks, use intermediate steps to reason, and intelligently decide when to retry or progress.
Orchestrates robot activities, like a high-level brain. Excels at planning and making logical decisions within a physical environment. Interacts in natural language, estimates progress, and can natively call tools – like using Google Search to look for information.
Uses precision pointing for spatial identification, motion reasoning, and safely handling objects under strict physical constraints.
Understands relationships across multiple camera streams to detect task success, and combines agentic vision with code execution to read complex industrial instruments.
Benchmark results comparing Gemini Robotics-ER 1.6 with Gemini Robotics-ER 1.5 and Gemini 3.0 Flash models. The instrument reading evaluations were run with agentic vision enabled (except for Gemini Robotics-ER 1.5 which doesn’t support it). All other evals were run with agentic vision disabled. The single view and multiview success detection evaluations contain different examples so are not comparable.
How the different elements of Gemini Robotics-ER 1.6 contribute to reaching a high level of performance on the instrument reading task.
Gemini Robotics-ER 1.6 improves substantially compared to Gemini Robotics-ER 1.5 on Safety Instruction Following which tests the ability to adhere to physical safety constraints. It improves compared to Gemini 3.0 Flash on pointing, and both models have very high accuracy for text. Gemini 3.0 Flash does better on bounding boxes.