Breaking

Google Positions Gemini Robotics as a Cross-Robot Intelligence Layer Rather Than a Single Humanoid Bet

October 4, 2026 at 14:00 EDT

Foundation Models
AI Agents
Robotics

Google is pushing Gemini beyond chatbots and into robotics, positioning its model family as a shared intelligence layer for many robot bodies rather than a bet on one flagship humanoid. The strategy centers on Gemini Robotics 2 and Gemini Robotics ER 2, a pair of systems meant to connect language, vision, planning and physical action across humanoids, tabletop robots and multi-robot workflows.

Oct. 4, 2026 · Google DeepMind

Gemini moves into the robot stack

Google is positioning Gemini Robotics as a shared intelligence layer for many robot bodies, from tabletop systems to full humanoids, rather than a single flagship humanoid product.

degrees of freedom in a five-finger hand

91.3%

moment-finding accuracy for ER 2

57.4%

progress classification accuracy

execution speed cited for moment finding

Moment finding versus progress tracking

91.3%

Moment finding

57.4%

Progress tracking

The strongest reported ER 2 metric is precise timing: knowing the critical moment to stop, switch tasks or verify success.

The robotics layer Google is assembling

Video and language

Understand instructions and live scenes.

→

Embodied reasoning

Plan steps, track progress and call tools.

→

Robot action

Hand off to VLAs, grippers, hands and APIs.

Why it matters

A shared model layer could let hardware teams use Gemini for reasoning, orchestration and action without building a full AI stack from scratch.

What remains hard

Google still cites limits in multi-finger dexterity, out-of-distribution tasks, high-degree-of-freedom control and broader access.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.