Google launches a tiny board that runs Gemma 3 locally

Google unveiled the new Coral Board at Google I/O - a compact single-board computer for on-device AI. The board features the Coral NPU, an open-source machine learning unit built on the RISC-V architecture and developed by Google Research. It's designed for small devices like headphones, AR glasses, and smartwatches, and aims to fix the fragmentation problem among AI accelerators.

At its core sits a Synaptics Astra SL2619 chip with a 2 GHz dual-core processor, 2 GB of RAM, and 1 TOPS of compute. Google's slightly older open-source language model Gemma 3 270M runs entirely on the board - no cloud needed. At I/O, Google showed demos including real-time translation, voice-controlled hardware, and a generative music performance where a YOLOv8 model tracked jellyfish movements and turned them into music. All demos are open source on GitHub. The board is expected to ship this summer, but Google hasn't announced a price yet.