Reading: Google launches Gemini 4 Neon robotics plugin on Google Antigravity

Google launches Gemini 4 Neon robotics plugin on Google Antigravity

Published
3 min read
Advertisement

Google launched the Robotics plugin on Google Antigravity and paired it with Gemini 4 Hydrogen, a faster model built to speed up robotics work without giving up direct robot-arm control. The new plugin adds hardware onboarding, simulation and robot control in one place, pushing developers from setup to a working application in a single session.

That is why Gemini 4 Neon is drawing attention now: the launch is aimed at the people who want to bring a real machine online fast, not just test ideas in a simulator. Google said the workflow can take a robot from boxed hardware to a working application in an afternoon, and that an arm can be calibrated and ready for tasks in about 10 minutes once the plugin is used.

The practical part matters because the plugin does more than open a screen and wait for a prompt. It can install software, find the camera, calibrate the arm and run test motions. In one setup, an I2RT YAM arm and an external depth camera can be brought up from an empty desk with Hydrogen in Antigravity. A similar flow can also bring a low-cost SO-ARM101 arm and an industrial grade FR3 Duo from scratch to live waving, which is the kind of result that turns a demo into a working development path.

- Advertisement -

Gemini 4 Hydrogen itself is built around speed and efficiency, but it still keeps the ability to drive a robot arm directly from a camera view and a plain-language goal. On the Artificial Analysis Robotics 0.1 benchmark, it controls a real 6-axis arm through eight everyday tabletop tasks and scores 58. That score is close to Gemini 4 Argon, while the cost per token is lower, and on public ERQA it also outperforms Gemini Robotics ER 2 on spatial reasoning tasks.

That creates the pressure point inside the launch. A faster model that is cheaper per token and stronger on several robot tasks sounds like the obvious answer, yet the same rollout keeps a separate Robotics plugin in the picture. Google is folding specialized skills from Gemini Robotics ER 2 into the core Gemini 4 series, but it is also preserving a separate tool layer for setup, simulation and control. In other words, the model is not replacing the workflow so much as making the workflow practical enough to use every day.

MuJoCo runs directly inside Antigravity as part of that workflow, with dedicated creation, auditing and rendering skills, so developers can build and preview high-quality simulated scenes and assets alongside their code. The launch marks a clear shift for robotics developers: not just a better model, but a faster route from empty desk to live hardware, with the remaining question being how far that combined stack can carry real-world robotics work beyond the first afternoon.

Advertisement
Share This Article