Jump to Content
Search Close
Search Close

Gemini Robotics-ER

Gemini Robotics

Our advanced Gemini 2.0-based models designed for the next generation of helpful robots

Join the trusted tester waitlist

Gemini Robotics brings Gemini’s capacity for multimodal reasoning and world understanding into the physical world - allowing robots of any shape and size to perform a wide range of real-world tasks.

Read the blog

Key capabilities

Gemini models are capable of responding to text, images, audio, and video. Gemini Robotics adds the ability to reason about physical spaces – allowing robots to take action in the real world.

  • Generality

    Uses Gemini's world understanding to generalize to novel situations, including dealing with new objects, diverse instructions, and new environments.

  • Interactivity

    Understands and responds to everyday commands, and reacts to sudden changes in instructions – or its surroundings. Then carries on without further input.

  • Dexterity

    Enables robots to tackle complex tasks requiring fine motor skills and precise manipulation – like folding origami, packing a lunch box, or preparing a salad.

  • Multiple embodiments

    Adapts to a diverse array of robot forms, from bi-arm robotic platforms like ALOHA 2 to humanoid robots like Apptronik’s Apollo.

Hands-on

Watch how Gemini Robotics performs different tasks.

Generality in action

Gemini Robotics can generalize across a vast range of tasks.

Dynamic interactions

Gemini Robotics responds to natural conversation and adapts rapidly to changing environments.

Partnering with Apptronik

Gemini Robotics is helping build the next generation of humanoid robots.

Dexterous skills

Gemini Robotics can perform tasks that require fine motor skills and coordination.

Responsibly advancing AI and robotics

To ensure Gemini Robotics benefits humanity, we’ve taken a comprehensive approach to safety, from practical safeguards to collaborations with experts, policymakers, and our Responsibility and Safety Council.

Learn more

Gemini Robotics model family

Models created to support embodied AI capabilities.

Collaborations

We’re partnering with Apptronik to build the next generation of humanoid robots. We’re also working with a select number of trusted testers to guide the future of Gemini Robotics-ER.

Experience
Gemini Robotics

If you're interested in testing our models, please share a few details to join the waitlist.

Join the trusted tester waitlist

Get the latest updates

Sign up for news on the latest innovations from Google DeepMind.

Explore our other teams and product areas

Gemini Robotics-ER

Gemini Robotics-ER

Our advanced embodied reasoning model

Our Gemini 2.0-based model gives advanced world understanding to robots.

View tech report

Key capabilities

  • Object detection

    Identifies and tracks the location and size of objects within 2D and 3D spaces.

  • Pointing

    Identifies objects and elements within those objects, to interact with them.

  • Grasp prediction

    Calculates how to grip objects, adjusting as necessary.

  • Trajectory reasoning

    Generates a plan of the necessary actions needed to complete a task.

  • Multi-view correspondence

    Can reason in 3D space and identify objects from different points of view.

Benchmarks

Capability Benchmark Description
Gemini 2.0 Flash
Gemini Robotics-ER
Pointing
Paco-LVIS
Localization of object parts across image (LVIS) and video (Ego4D) datasets 46.1% 71.3%
Pixmo-Point
Challenging pointing dataset, requires accurate localization of objects in cluttered scenes. 25.8% 49.5%
Where2Place
Predict spatial localization that requires one-step reasoning. 33.8% 45%
3D detection
SUN-RGBD
Predict object 3D bounding boxes from monocular RGB images in SUN-RGBD indoor scenes dataset. 30.7% 48.3%

Model information

Model deployment status Private preview
Supported data types for input Image, Video, Text
Supported data types for output Text
Supported # tokens for input 32k
Knowledge cutoff June 2024
Availability Trusted tester