Google has unveiled its latest innovation in robotics AI — the Gemini Robotics On-Device model — designed exclusively for robots, not humans. The tech giants describe the it as their most powerful VLA (vision language action) model that is optimised to run locally on robotic devices.

What is Gemini Robotics On-Device?

Gemini Robotics On-Device is a robotics foundation model, engineered specifically for bi-arm robots. Unlike many AI models that rely on cloud connectivity, this model operates independently of data networks, it’s helpful for latency sensitive applications, and environments where connectivity is limited or non-existent.

The model exhibits strong general-purpose dexterity and task generalisation, enabling robots to perform complex, real-world tasks with increased precision and adaptability — all while running efficiently on the robot itself.

Gemini Robotics On-Device: Key Features and Performance

According to Google, Gemini Robotics On-Device is – 

• Designed for rapid experimentation with dexterous manipulation.
• Adaptable to new tasks through fine-tuning to improve performance.
• Optimized to run locally with low-latency inference.
• Gemini Robotics On-Device achieves strong visual, semantic and behavioral generalization across a wide range of testing scenarios, follows natural language instructions, and completes highly-dexterous tasks like unzipping bags or folding clothes — all while operating directly on the robot.

On a separate note, Google has also added an AI mode feature to India as an experiment in Labs in English, ANI reported. This feature will provide more advanced reasoning and multimodality, as well as the ability to go deeper through follow-up questions and helpful links to the web.

“Using a custom version of Gemini 2.5 lets you ask longer, more complex or nuanced questions that would have previously required multiple searches,” Google said.

Early tester of this AI mode lets you ask much longer queries, two to three times the length of traditional searches. This would be helpful for exploratory questions and for more complicated tasks like comparing products, planning a trip, or understanding complex how-tos.

This AI mode feature is multimodal, which means that one can speak their question or even snap a photo to raise a query.

Google believes this launch is part of its long-term vision to make finding and accessing information even more effortless in Search. It further adds that this would help people to turn to Google Search to tackle increasingly complex and nuanced questions.

“AI Overviews, now used by over 1.5 billion users monthly, is a testament to this evolution. In our biggest markets like the U.S. and India, it’s already driving a more than 10% increase in usage for the types of queries where they appear,” Google said.

“This implies that once people use AI Overviews, they’re coming to do more of these types of queries, and what’s particularly exciting is how this growth increases over time,” Google added.

With the inputs of ANI