Google DeepMind запустила Gemini Robotics 2 с полным контролем тела робота

Google DeepMind announced the release of Gemini Robotics 2 — an updated model for robot control that, according to the company, for the first time provides full control over the entire body of a machine and allows it to reason about every movement in real time.

What Changed in Gemini Robotics 2

The key innovation of Gemini Robotics 2 is full-body control. Traditional robot control systems pre-program individual movements in advance: the arm rises according to one algorithm, the leg takes a step according to another. This approach works in predictable conditions but fails at the slightest change in the environment.

According to Google DeepMind, Gemini Robotics 2 applies a different logic: the model reasons about each movement at the moment it is performed — on a principle close to how a human walks on an uneven surface or grasps an object of an unfamiliar shape. The system coordinates all parts of the body simultaneously rather than controlling them sequentially.

• Developer: Google DeepMind
• Model: Gemini Robotics 2
• Key feature: full-body control
• Principle: reasoning about every movement in real time
• Result: a wide range of physical tasks previously inaccessible to robots

Why This Matters for Robotics

Full-body control is one of the key unsolved challenges in the industry. Most industrial robots handle repetitive movements on production lines well but become lost in non-standard situations: a fallen object, an unfamiliar container, a changed route.

Humanoid robots — actively being developed by Tesla, Figure AI, Boston Dynamics, and other companies — require exactly this ability: to dynamically coordinate the entire body in real time. If Gemini Robotics 2 solves this task at the level of a foundational model, it could potentially accelerate the training of robots on new tasks without separately programming each movement.

"Gemini

Robotics 2 allows robots to reason about every movement, opening up a wide range of tasks," states the official announcement from Google DeepMind, as quoted by The Robot Report.

Context: Gemini in Physical AI

Google DeepMind has been consistently advancing the application of its multimodal models in robotics. The first version of Gemini Robotics was focused on perception and basic manipulation. Gemini Robotics 2, judging by its description, takes the next step — from recognition to active body control.

The direction of Physical AI in 2025–2026 has become one of the most competitive in the industry: OpenAI, Google, Meta, and dozens of startups are increasing investments in robots capable of operating in unstructured environments — in warehouses, hospitals, and domestic settings.

What This Means

If Gemini Robotics 2 genuinely provides general reasoning about body movements, this is a serious claim by Google DeepMind to leadership in Physical AI. For the industry, this means that foundational models can take on tasks that previously required specialized programming for each type of robot — and each specific task.

*Meta has been recognized as an extremist organization and is banned in Russia.