Google DeepMind发布 Gemini Robotics 2，推进机器人全身智能控制

Google is tearing down its old AI structure to build something broader, pushing hard into whole-body robotics and desktop tools. DeepMind just launched Gemini Robotics 2, describing it as "one brain for any robot." The setup splits into three parts: a vision-language-action model for full-body movement, Gemini Robotics ER 2 for multi-robot planning, and an on-device model that adapts to new hardware in hours using under 200 examples. To keep these physical agents in check, Google also introduced ASIMOV-Agentic, a safety benchmark designed to measure if a robot knows when to stop and ask a human for help. That same automation push is overhauling Google’s software ecosystem.

We demonstrated how Gemini's multimodal understanding could drive real-world action with Gemini Robotics. Today, we are introducing Gemini Robotics 2 - the intelligence layer powering the next generation of truly adaptable robots. As it takes its first literal steps, this major advance unlocks intelligent whole-body control, advanced dexterity, and multi-robot collaboration. [...] Gemini Robotics 2: Our most advanced vision-language-action model (VLA) that converts vision and language input into motor control, enabling a robot to take action. This model is capable of controlling full humanoids, from feet to fingertips, and other bi-arm robots. It also brings a new level of dexterous manipulation on both hands and grippers. Gemini Robotics ER 2: Our most capable embodied reasoning (ER) model. It is a vision language model (VLM) that acts as our agent, enabling robots to communicate with humans, understand the physical world and plan multi-step tasks lasting several minutes. We are also

# Gemini Robotics 2 Brings Whole-Body Intelligence to Robots Humanoid robot carrying a watering can while a dual-arm robot works at a lab bench, illustrating Gemini Robotics 2 whole-body intelligence Last updated: July 30, 2026 Last verified: July 30, 2026 TL;DR: Gemini Robotics 2 is Google DeepMind’s new three-model robotics stack. It adds whole-body humanoid control, longer task planning, multi-robot coordination and faster on-device adaptation. The technical progress is real, but the strongest evidence still comes from Google’s controlled demonstrations and internal evaluations. Multi-finger dexterity is inconsistent, the action models remain restricted, and Google’s own safety report says its tests do not replace certified hardware and functional-safety systems. [...] The upgrade is broader than “better hands.” Google is connecting perception, long-horizon planning, motor control and fleet orchestration into one family. ## What this release means for physical AI Gemini Robot

One brain. For any robot. 🤖 Meet Gemini Robotics 2: our next-generation AI model bringing intelligent whole body control, advanced dexterity, and multi-robot collaboration to embodiments of different shapes and sizes. Three new models drive this breakthrough: 1️⃣ Gemini Robotics 2: A vision, language, and action model controlling full humanoids from feet to fingertips. 2️⃣ ER 2: A reasoning agent built to plan complex, end-to-end tasks lasting several minutes. 3️⃣ On-Device 2: Runs locally and adapts to new robot bodies in just a few hours. [...] Ahmad Abbous, graphic “One brain. For any robot.” is the headline. The part I’m watching is where embodiment stops being an interface detail—different joints, failure modes, and safety margins can make the same plan a different kind of risk. Ievgenii Tsokalo, graphic #mimetik digitizes manual work with high fidelity, turning top engineers’ skills into reusable, machine-readable knowledge. This data can be used by Gemini Robotics 2 to deplo

Introducing Gemini Robotics 2 - the intelligence layer powering the next generation of truly adaptable robots. As it takes its first literal steps, this major advance unlocks intelligent whole-body control, advanced dexterity, and multi-robot collaboration. Learn more: ___ Subscribe to our channel Find us on X Follow us on Instagram Add us on Linkedin [...] bag task, in particular, several people did think it was impossible. You don't think about how to drive 22 separate joints when you operate your hand, but that's what we're asking these AI models to do. And third thing is, in Gemini Robotics 2 we add a feature that enables multiple robots to collaborate to accomplish the same task simultaneously. [Apollo] Hey Duo, kit all tools in the bin, close the kit and put the kit back into the bin. [Duo] On it. Each of these two robots is running that same stack, and so rather than having one neural network that controls both robots each have their own copy, and they're doing their ow

Log inSign up ## Post user avatar Google DeepMind @GoogleDeepMind Jul 30 One brain. For any robot. 🤖 We’re launching Gemini Robotics 2: our next-generation physical AI bringing full body intelligence to humanoids, advanced dexterity, multi-robot teamwork and more. 00:00 1.6M user avatar Google DeepMind @GoogleDeepMind Jul 30 Three new models power this breakthrough: 1️⃣ Gemini Robotics 2: A vision-language-action model controlling humanoids from feet to fingertips 2️⃣ Gemini Robotics ER 2: Capable of real-world video understanding and complex, multi-step planning 3️⃣ On-Device 2: Runs locally and adapts 00:00 39K user avatar Google DeepMind @GoogleDeepMind Jul 30 [...] Google DeepMind @GoogleDeepMind Jul 30 Gemini Robotics 2 moves physical AI beyond tabletop tasks, enabling intelligent whole-body control for humanoids. Watch @Apptronik's Apollo 2 process a single prompt to reach, bend, and pick up a watering can ↓