Google AI Edge Gallery
Run small language models locally on phones and computers
Tiny special-purpose models that run locally on everyday devices
A family of very small AI models designed to run on-device rather than in the cloud, aimed at makers building for Raspberry Pis, wearables, smart-home hubs and mobile apps. It's for narrow, high-volume tasks where calling a cloud API is wasteful or too slow.
You run the compact model locally through the project's runtime, so inference happens on the device with no per-request cost or network round trip. It works well for extracting or classifying structured data from documents and messages where quality-per-token matters more than general reasoning.
Top 10 GitHub: AI videos, gorgeous diagrams, token savings and more
Run small language models locally on phones and computers
Train a tiny language model from scratch on regular hardware
Python-like language with C-level performance for AI code
Plugin-based local harness for private, subscription-free AI work
Debloat Android TV and Google TV over Wi-Fi
Disguises proxy traffic as ordinary website visits