Image Courtesy: Google

Google has announced the release of Gemini Robotics-ER 1.5, its most advanced robotics reasoning model to date—and for the first time, it’s available to all developers. This marks a major step in making state-of-the-art embodied reasoning accessible to the broader robotics and developer community.

What Makes Gemini Robotics-ER 1.5 Unique?

Unlike traditional models, Gemini Robotics-ER 1.5 is designed as a high-level reasoning “brain” for robots. It combines advanced spatial and visual understanding with task planning and progress estimation, enabling robots to take on complex, real-world challenges.

For example, imagine asking a robot:
"Can you sort these objects into compost, recycling, and trash bins?"

To succeed, the robot must:

• Check local recycling rules online,

• Identify the items in front of it,

• Plan the sorting steps,

• And physically carry out the task.

Gemini Robotics-ER 1.5 is built to handle this kind of multi-step, context-heavy workflow with strong performance across academic benchmarks and real-world test cases.

Key Features in Gemini Robotics-ER 1.5

• Fast, powerful spatial reasoning: Low-latency performance similar to Gemini Flash, with precise 2D point generation grounded in reasoning about size, weight, and affordances.

• Advanced agentic behaviors: Reliable long-horizon planning, progress tracking, and the ability to call external tools like Google Search or third-party APIs for context-driven execution.

• Flexible thinking budget: Developers can balance speed and accuracy—letting the model “think longer” for complex tasks or respond instantly for reactive ones.

• Improved safety filters: Enhanced semantic safety prevents unsafe or impractical plans (e.g., lifting objects beyond a robot’s capacity).

The Agentic Brain for Robots

Google positions Gemini Robotics-ER 1.5 as the decision-making and orchestration layer for robotics systems. It can break down natural language instructions like “clean up the table” into executable plans, seamlessly calling hardware APIs, grasping models, or vision-language-action (VLA) models to complete the task.

Availability

Gemini Robotics-ER 1.5 is now in preview via Google AI Studio and the Gemini API. Developers can start building robotics applications that harness embodied reasoning, bridging the gap between perception, planning, and action.

Know more.