Google DeepMind Showcases Gemini Robotics 1.5, Watch Robots Learn, Plan, And Perform Seamlessly

Google DeepMind has unveiled two new artificial intelligence (AI) models in its Gemini Robotics family, designed to make robots smarter and more capable in the real world. Named Gemini Robotics-ER 1.5 and Gemini Robotics 1.5, the models work together to improve a robot’s reasoning, vision, and action.

Unlike earlier AI systems that combined planning and execution in one model, DeepMind’s new approach separates these tasks. Gemini Robotics-ER 1.5 acts as the planner or orchestrator, while Gemini Robotics 1.5 carries out tasks based on natural language instructions. This two-model system aims to reduce errors and delays often seen when a single AI handles both planning and performing actions.

Gemini Robotics-ER 1.5 is a vision-language model (VLM) that can analyse a situation, create multi-step plans, and integrate external tools such as Google Search to make informed decisions. It excels in spatial understanding, allowing robots to navigate and manipulate their surroundings more effectively.

Once a plan is ready, Gemini Robotics 1.5, a vision-language-action (VLA) model, translates the instructions and visual information into motor commands. This enables robots to complete tasks efficiently while also explaining their actions in natural language. The models together allow robots to handle complex tasks step by step. For instance, a robot could sort items into compost, recycling, and trash bins by checking local recycling rules online.

DeepMind says the system is adaptable to robots of various shapes and sizes thanks to its spatial awareness and flexible design. Currently, developers can access the planner model, Gemini Robotics-ER 1.5, through the Gemini API in Google AI Studio. The task-executing model, Gemini Robotics 1.5, is available only to select partners.

ALSO READ: Simple Google Gemini Nano Banana AI Tips To Turn Your Selfie Into Stunning Diwali, Garba And Puja Looks