Instead of sending every request to a distant server, edge AI runs the model directly on a phone, camera or appliance. Responses come back instantly, work without a connection, and the data, which may be personal, need never leave the device.
The constraint is size: the hardware is modest, so models must be compact and efficient, often shrunk through distillation or quantization. The pay-off is privacy and responsiveness, which is why more everyday devices now carry capable models of their own.