The maturity of Internet of Things technology and the widespread deployment of network provide good infrastructure conditions to the application of AI at the device, edge and cloud for IoT and smart city. The requirements of AI models in IoT and smart city are dynamic, thus how to make the AI system continuously and dynamically update, as well as how to infer and predict in real-time, is essential to the application of AI in IoT and smart city. The device-edge-cloud collaborative service enables collaborative inference, and dynamic learning and updating of AI models on the device-edge-cloud architecture, so as to meet the needs of various current and future application scenarios. This recommendation specifies the functional architecture of AI enabled device-edge-cloud collaborative services for IoT and smart city.