For Developers/Models/Gemini Robotics ER 2
Closed SourcePreviewGoogle DeepMindReleased 2026-07

Gemini Robotics ER 2

High-level reasoning brain for robots enabling video understanding, task orchestration, and multi-robot collaboration

Context window

128K

Input / 1M tokens

Not announced

Output / 1M tokens

Not announced

Provider

Google DeepMind

Data verified 2026-07-31

Gemini Robotics ER 2 is an embodied reasoning model that serves as a high-level brain for robots. It enables real-time spatial reasoning, multi-step task planning, and collaboration between different robots while monitoring progress through video understanding.

Capability index

Relative estimates (0-100) to place this model against its peers, grounded in published benchmarks.

Coding
0
Reasoning
9
Math
0
Multimodal
10
Long context
7
Speed
7
Cost efficiency
0

How to access it

Available via Gemini API, Google AI Studio, and Gemini Enterprise Agent Platform for developers building physical AI agents.

Strengths

  • Advanced video understanding and progress tracking
  • Multi-robot collaboration capabilities
  • Long task sequence execution (minutes-long, hundreds of decisions)
  • Real-time streaming via Live API
  • Natural language interaction

Best for developers who...

Complex multi-step robotic tasksMulti-robot coordination and collaborationTasks requiring progress monitoring and temporal understandingNatural language-based robot control

When to choose it (and when not to)

Reach for Gemini Robotics ER 2 when...

  • When robots need to understand task progress and adapt in real-time
  • For orchestrating multiple robots on shared workflows

Look elsewhere if...

  • Low-latency on-device execution (use On-Device 2 variant)
  • Direct motor control (pairs with VLA models)

How to use it

  • Use natural language instructions with task context
  • Leverage video input for progress tracking
  • Define multi-step workflows as sequential instructions

Quickstart

Python
from google.genai import Client
client = Client()
response = client.models.generate_content(
    model='gemini-robotics-er-2-preview',
    contents=['Plan steps to clean this room'])

Requires Gemini API key; streaming available via Live API

API model id: gemini-robotics-er-2-preview

Benchmarks

BenchmarkScoreNotes
Progress Classification Accuracy57.4%Tracking task progress in five levels (0-100%)
Moment Finding Accuracy91.3%Identifying exact video frame of critical events with 0.96s MAE
Unscrewing Light Bulb Task92%Success rate on full-body coordination task

Source: Google DeepMind Official Announcement

Compare Gemini Robotics ER 2

Compare Gemini Robotics ER 2 with any other model

Build a comparison →
All model comparisons →

Learn the concepts