Google has introduced a 3D-printable Open Duck robot powered by its Gemma 4 technology. This initiative leverages the open-weight models of Gemma 4, which are specifically designed for on-device applications like robotics and agentic AI. The smaller effective-parameter variants of these models are optimized for low-latency performance on consumer-grade hardware, ideal for hobbyist and educational robots. This unveiling emphasizes practical demonstrations of Gemma 4’s capabilities in real-world robotics scenarios.

Google: Google is a global technology company focused on internet services, cloud computing, hardware, and artificial intelligence, with its Google DeepMind division developing the Gemma family of open-weight AI models. In this news, Google is demonstrating the Gemma 4 model running on an Open Duck robot platform, highlighting how its open, on-device AI can power low-cost, user-assembled robotics experiences.
Open Duck: Open Duck is a Gemma 4-powered open-source robot platform that Google is showcasing as a 3D-printable, user-assembled hardware companion for its local AI models. In this context, Open Duck serves as a tangible demo of how Gemma 4 can enable agentic, multimodal behavior on affordable, customizable robotic hardware that users can build themselves.

`json
{
“Gemma_4_Open_Models”: “Google DeepMind’s Gemma 4 family is released as open-weight models under a permissive license, designed for on-device and edge deployments including robotics and agentic AI use cases.”,
“On_Device_Agentic_AI”: “Gemma 4’s smaller effective-parameter variants are optimized for low-latency inference on consumer-grade hardware, suitable for responsive, offline control of hobbyist and educational robots.”,
“Robotics_Demonstrations”: “Recent Gemma 4 launch materials emphasize physical-world demos such as simple home or desk robots to show how the models’ multimodal perception and tool-calling can be applied to real-world actions.”
}
`