Google DeepMind Launches Gemini Robotics On-Device for Offline AI Control

Google DeepMind unveiled Gemini Robotics On-Device, a new vision-language-action model designed to operate locally on robots without requiring an internet connection. The model builds on March's Gemini Robotics AI release, offering enhanced efficiency and adaptability for real-time robotic control.

Carolina Parada, Google DeepMind's Head of Robotics, confirmed the model's hybrid predecessor combined cloud and on-device processing. The latest iteration delivers near-flagship performance offline, with developers able to fine-tune operations using natural language prompts.

Benchmark tests show the compact model rivals its cloud-based counterpart in capability. This advancement signals Google's push toward decentralized AI solutions in robotics, though no direct cryptocurrency implications were noted in the announcement.