On July 23, the Beijing Municipal Development and Reform Commission, along with other departments, jointly issued the "Several Measures for Beijing to Accelerate Development Led by Intelligent Agents," aiming to drive innovation and development in the field of intelligent agents and cultivate new forms of the intelligent economy. The measures encourage the development of a Token economy.

They call for innovation entities to develop general-purpose processors suitable for intelligent agent system calls, complex task scheduling, and high-frequency decision-making, as well as to create specialized inference chips with low latency and high throughput.

The initiative also promotes the tackling of key technologies, such as inference architecture, by innovation entities. It aims to reduce inference costs through heterogeneous collaboration, memory-computation synergy, and intelligent scheduling, while accelerating efficiency optimizations at the application layer, including inference cache reuse and intelligent task routing. The goal is to optimize the entire chain to improve Token efficiency.

The measures seek to reshape product forms and service models, fostering new business paradigms like Token-as-a-Service (TaaS), Agent-as-a-Service (AaaS), and Result-as-a-Service (RaaS). They also aim to include more eligible new Token products and services within the scope of product catalogs supported by service vouchers for small and medium-sized enterprises.

Support will be increased through measures like computing power vouchers, and districts with suitable conditions are encouraged to collaborate with innovation entities to explore the issuance of Token vouchers and intelligent agent service vouchers to accelerate the promotion and application of intelligent agents.

Financial institutions, including banks and insurance companies, are supported in developing various financial products that comply with regulations to facilitate the implementation of intelligent agents.

Innovation entities are also supported in legally and compliantly aggregating industry knowledge and intelligent agent execution data for open sharing, building a data flywheel to continuously enhance the capabilities of intelligent agents.