Google DeepMind shared a major update in robotics. Google has introduced Gemini Robotics On-Device, their most advanced model yet, which brings AI directly to local robotic devices. The VLA model is optimised to run locally on robotic devices to handle general-purpose dexterity and task management. Google said, “In our evaluations, our On-Device mode exhibits strong generalisation performance while running entirely locally. Gemini Robotics On-Device also outperforms other on-device alternatives on more challenging out-of-distribution tasks and complex multi-step instructions.” Gemini Robotics On-Device is a lightweight robotics foundation model made for bi-arm robots. It supports quick testing, adapts to new tasks with fine-tuning, and runs locally with low delay for fast response. Google is offering a Gemini Robotics SDK for developers to test the On-Device model in different tasks and environments. It works with the MuJoCo simulator, and developers can access it by joining the tester programme. Google AI Mode: Google Introduces Advanced Multimodal AI Search in India, Powered by Gemini 2.5. Gemini Robotics On-Device Model We’re bringing powerful AI directly onto robots with Gemini Robotics On-Device. 🤖 It’s our first vision-language-action model to help make robots faster, highly efficient, and adaptable to new tasks and environments – without needing a constant internet connection. 🧵 pic.twitter.com/1Y21D3cF5t — Google DeepMind (@GoogleDeepMind) June 24, 2025