Illustration for: Google DeepMind Unveils Gemini Robotics 2 for Whole-Body Control

Google DeepMind Unveils Gemini Robotics 2 for Whole-Body Control

Google DeepMind introduced Gemini Robotics 2, a three-model suite giving robots whole-body control -- walking, crouching and manipulating objects while reasoning through tasks -- beyond the prior model's upper-body-only control.

By the Numbers

3 (VLA, ER, on-device)
Model components
Whole-body
Control scope
Upper-body only
Prior model scope
TC
By the AI Desk
Edited by Trace Cohen · Early-stage VC & angel · Founder, New York Venture Partners
1 min read
ShareXLinkedInEmail

THE RUNDOWN

1

Gemini Robotics 2 includes three components: a vision-language-action model that converts visual and language inputs into full-body robot movement, Gemini Robotics ER 2 for multi-step task planning and multi-robot coordination, and an on-device version optimized to run locally on robotic hardware

2

Unlike its predecessor, which primarily controlled a robot's upper body, the new system can direct an entire humanoid from top to bottom -- walking, crouching, and manipulating objects while reasoning through the task at hand

3

Google released demonstration footage showing tasks including cleaning up trash, picking up watering cans, inserting a tape into a boombox, screwing in a lightbulb, and tying a garbage bag closed

4

The release positions Google alongside a small group of labs -- including those building on Nvidia's robotics stack -- racing to solve general-purpose physical-world manipulation, a capability gap that has lagged well behind language and reasoning progress

TC

The VC Read · Trace's Take

Trace Cohen

Whole-body control sounds incremental until you realize it's the difference between a robot that can gesture and one that can actually do a warehouse shift. Physical-world manipulation has been the stubborn laggard behind language-model progress for years, and Google throwing DeepMind's full weight at it is a signal the capability gap is finally starting to close -- the question for robotics startups is whether Gemini Robotics ships as a platform they can build on, or as Google's own moat against them.

Analysis

Google DeepMind unveiled Gemini Robotics 2, a three-model suite that gives robots whole-body control -- walking, crouching and manipulating objects while reasoning through multi-step tasks -- a meaningful expansion beyond its predecessor, which primarily controlled a robot's upper body and arms.

The suite includes a vision-language-action model that converts visual and language inputs into full-body robot movement across both humanoid and dual-arm systems; Gemini Robotics ER 2, an embodied-reasoning model that plans multi-step tasks, communicates with people, and coordinates multiple robots working together; and an on-device variant optimized to run locally on robotic hardware without a constant cloud connection. Google released demonstration footage showing the system performing everyday tasks including cleaning up trash, picking up a watering can, inserting a tape into a boombox, screwing in a lightbulb, and tying a garbage bag closed.

Prior-generation systems that only handled upper-body control were effectively limited to a narrower band of tasks regardless of how capable their reasoning was.

The jump from upper-body to whole-body control matters because most commercially useful robotic tasks -- warehouse work, home assistance, light manufacturing -- require coordinated movement across a robot's entire frame, not just arm-level manipulation. Prior-generation systems that only handled upper-body control were effectively limited to a narrower band of tasks regardless of how capable their reasoning was.

For robotics and AI investors, Gemini Robotics 2 puts Google in more direct competition with the growing ecosystem of humanoid robotics startups and Nvidia's own robotics stack, all racing to close a capability gap that has lagged well behind progress in language and reasoning models -- physical-world manipulation remains genuinely harder to generalize than text or code generation. What to watch: whether Gemini Robotics 2 ships as licensable infrastructure for third-party robot makers, similar to how Google licenses other AI capabilities, or remains primarily an internal and Alphabet-affiliated robotics research showcase.

ShareXLinkedInEmail

Key Sources

2 sources

Reported by Bloomberg · Analysis by Value Add Pulse.

← Back to Pulse

THE WIRE in your inbox— Tech, startup & VC news with Trace's take. Free, no spam.