From foundational models to framework performance, inference efficiency, and industry applications, Beijing has been steadily advancing over the past six months, driven by intelligent agents. As a leader in China's AI development, Beijing has consistently taken the initiative, but it also faces new challenges such as a shortage of high-performance computing power, a lack of standards for scaling intelligent agent deployment, and the need for improved service systems.
In response, the Beijing Municipal Development and Reform Commission, in collaboration with relevant municipal departments, formulated and released on July 23 the "Several Measures for Beijing to Accelerate the Leading Development of Intelligent Agents" (hereinafter referred to as the "Intelligent Agent Measures"). This policy covers four areas: technology, application, industry, and ecosystem, with a total of ten specific measures. According to data from the China Business Industry Research Institute, the global AI intelligent agent market is estimated to be worth approximately $11.3 billion in 2025, and this figure is expected to reach $17.5 billion in 2026.
The Beijing Municipal Development and Reform Commission will dynamically reserve and implement a series of key projects through mechanisms such as "open competition" and "horse racing," and will provide up to 100 million yuan in financial support to selected projects. In the first half of 2026, the total financing scale in the city's AI sector exceeded 95 billion yuan, accounting for over 30% of the national total, and the number of newly listed companies has already matched the total for the entire previous year.
Targeting Breakthroughs in Control Layer Common Technologies
Technology is the foundation for the development of intelligent agents, and the Intelligent Agent Measures focus on two key aspects. "On one hand, we continue to push the boundaries of foundational models," said Lin Jianhua, Deputy Secretary of the Party Leadership Group and Deputy Director of the Beijing Municipal Development and Reform Commission, citing the "Reach Higher Plan" recently proposed by ZhiPu founder Tang Jie. "This is also a consensus in Beijing. We hope that elites from all sides will join in the ambitious goal of continuous advancement, supporting innovation in model architecture and training paradigms in areas such as online learning, continuous learning, and autonomous evolution, to constantly break through the ceiling of intelligence."
Lin also emphasized strengthening common technology breakthroughs in the intelligent agent control layer. The Intelligent Agent Measures specifically mention supporting the implementation of a control layer engineering project. This includes consolidating the common foundation of intelligent agents in areas such as context engineering optimization, task persistence, multi-agent collaboration, and system scalability. It also involves accelerating cross-model and cross-chip common intelligent agent technology to improve the length of complex task chains and the stability of long-term task execution. The goal is to build autonomous and controllable interconnection protocols, development frameworks, and basic tool chains, develop standards and specifications, support the creation of intelligent agent infrastructure like "Lingji," enhance the capability for intelligent agent development, deployment, and cross-framework interoperability, while ensuring safety and controllability.
In 2025, the Municipal Bureau of Economy and Information Technology released the "Action Plan for Beijing to Support Information and Software Enterprises in Enhancing AI Application Service Capabilities (2025)" (referred to as the "AI Software Eight Policy"). Now, the Intelligent Agent Measures have been released. Chen Zhaohui, Deputy Director of the Beijing Municipal Bureau of Economy and Information Technology, elaborated, "The AI Software Eight Policy released in 2025 primarily addressed the issue of 'how enterprises can effectively use AI.' The current Intelligent Agent Measures mainly address how the industry can build a robust intelligent agent ecosystem." Chen added, "The policy upgrade is reflected in three areas: the policy focus has shifted from AI empowering software to intelligent agents restructuring software; the scope of support has expanded from software companies to the entire industry chain; and the policy system has been upgraded from individual breakthroughs to systemic restructuring."
In terms of systemic restructuring, the Intelligent Agent Measures are a comprehensive policy framework. They introduce new concepts for the first time, such as an intelligent agent infrastructure base, an intelligent agent interconnection protocol (AIP), a Token economy, and the OPC (One Person Company) innovation and entrepreneurship model. From technological innovation and infrastructure development to application scenarios, new business incubation, safety governance, and resource guarantees, they form a policy system covering the entire lifecycle of intelligent agents.
Emphasis on Intelligent Agent Native Applications
Regarding applications, the Intelligent Agent Measures include measures to accelerate the deployment of intelligent agent native applications and to promote the integrated development of intelligent terminals and intelligent agents. For intelligent agent native applications, Lin Jianhua explained that the measures support the development of AI-native super software and artificial intelligence operating systems. They also support the creation of scientific AI assistants and platforms capable of information correlation and knowledge system building, the development of AI-native autonomous laboratories, and industry solutions that are data-driven, autonomously decision-making, and human-machine collaborative. The measures also promote models like the Frontier Deployment Engineer (FDE) to deepen industry application deployment and create more benchmark demonstration scenarios in fields such as science, healthcare, and manufacturing.
As early as 2024, Beijing was the first in the country to release an "AI+" action plan. In the next step, leveraging the deep evolution and widespread penetration of intelligent agents, the Beijing Municipal Development and Reform Commission will focus on breakthroughs in key areas, regions, data loops, common platforms, and technologies. Lin also mentioned, "We support the development of an AI OS with cross-device collaboration, cross-ecosystem development, and cross-scenario deployment capabilities to accelerate the application of intelligent agents in more terminal products. Last year, Beijing was the first to launch the Doubao phone. We will optimize inference engines to reduce costs and improve efficiency for intelligent agents and build intelligent agent development and service platforms to lower the barrier for developers."
Liu Weihua, a member of the Party Leadership Group and Deputy Director of the Beijing Municipal Science and Technology Commission and the Zhongguancun Science and Technology Park Administrative Committee, provided a detailed overview of intelligent agent applications in the science and technology field. "You can think of a research intelligent agent as an 'AI scientist.' Earlier, we continuously promoted the upgrade and iteration of the Panshi scientific foundational large model, just releasing version 2.0 last week. Based on this, we are building an integrated intelligent research platform to support researchers in quickly customizing specialized models and research intelligent agents on demand. Currently, multiple intelligent agents have been deployed on a large scale. For example, the Literature Compass intelligent agent can generate professional-level literature reviews within three hours with 90% accuracy. It has generated nearly 40,000 reviews and provided over a million scientific knowledge Q&A reasoning services."
Liu stated that in the next step, Beijing will accelerate core technology breakthroughs across the entire chain of research intelligent agents. Using autonomous laboratories as the core carrier, it will promote the efficient application of intelligent agents in scientific research while coordinating the full supply of computing power, data, and talent.
Supporting New Business Models like OPC and Token Economy
In the first half of 2026, Beijing saw the emergence of new, minimalist entrepreneurial models represented by OPC (AI One Person Company) and new business models like the Token economy. A group of industry companies achieved annualized revenue growth of several times, or even more than ten times. The Intelligent Agent Measures mention supporting new entrepreneurial models represented by OPC to cultivate new forms of corporate organization in the intelligent era. This involves building a layered and categorized professional service system and upgrading public services around flexible computing power supply and professional incubation. It also requires actively conducting pilot explorations to meet the access and operational needs of OPCs, making it easier for them to establish and grow.
The Intelligent Agent Measures also clarify the need to closely follow the intrinsic needs of the Token economy. This includes supporting the development of specialized inference chips, conducting key technology research in inference architecture, optimizing application layer efficiency like inference caching and reuse and intelligent task routing, and improving Token efficiency across the entire chain. It also involves cultivating new business models such as Token as a Service (TaaS), Intelligent Agent as a Service (AaaS), and Results as a Service (RaaS), and promoting the inclusion of more Token products and services in the scope of SME service voucher products. At the same time, it encourages a shift in Token billing from quantity to quality to enhance economic efficiency and improve the Token economic monitoring and evaluation methodology system.
Using the AI Origin Community as an example, Tang Chao, Deputy Director of the Zhongguancun Science City Management Committee and Vice District Mayor of Haidian District, stated, "The AI Origin Community has gathered over 1,000 AI scientists, more than 13,000 AI developers, and over 100,000 AI students. It serves over 30 universities and research institutions in the surrounding area and has attracted over 400 AI companies to settle in, forming a '10-minute innovation circle.'" Discussing achievements in gathering source innovation resources, connecting the full OPC cultivation chain, and providing diversified supporting facilities, Tang said, "Haidian District has set up exclusive open workstations for OPCs and launched the OPC 'Jiedanba' platform to supply application scenarios. It has gathered nearly 180 OPC companies, primarily composed of young people aged 25-30. Haidian District has also introduced a combined policy of computing power subsidies and Token vouchers to reduce the cost of computing power calls from both directions. Simultaneously, it has established a flexible computing power laboratory and collaborated with cloud service companies to build an inclusive computing power supply system."
For OPC growth and collaboration, the Beijing Economic-Technological Development Area (EDA) also has its own plans and achievements. According to Li Yantao, Deputy Director of the Beijing Economic-Technological Development Area Administrative Committee and Director of the Beijing Integrated Circuit Major Project Office, the EDA has built 8 OPC communities, including the city's first. It has released the country's first open-source intelligent agent common infrastructure, providing a secure and controllable runtime base for OPCs. Collaborating with leading domestic AI open-source communities, it has established a full-cycle growth service system for OPCs, offering one-stop services from technical exchange and resource connection to commercial deployment. It has innovatively built an OPC business collaboration mechanism and launched a super-individual intelligent order distribution platform, facilitating the completion of over one hundred high-quality orders.
Preventing Malicious Misuse of Technology
Security governance is a safeguard for the development of intelligent agents. The Intelligent Agent Measures clearly state the need to coordinate security governance with open-source and open development. This involves improving the governance capabilities of intelligent agents, further regulating the industry's development to prevent malicious misuse of technology, developing security large models to enhance the ability to "govern models with models," and accelerating the improvement of common service capabilities in areas such as intelligent agent vulnerability scanning, security testing, offensive and defensive confrontation, and security certification.
Li Degang, Deputy Director and Spokesperson of the Beijing Municipal Internet Information Office, elaborated on building an innovative, open, inclusive, and prudent governance model: "We will leverage Beijing's advantages in technological talent and the 'Two Zones' policy. Based on the already filed Xiaomi, Samsung, and Doubao phone assistants approved by the Central Cyberspace Administration, we will focus on commercial exploration of intelligent agents, including software-hardware integration and value assessment. We will fully implement the policy for managing AI-generated synthetic content labeling, implement 'explicit' and 'implicit' dual-labeling requirements, and regulate the positive interaction of AI-generated content. We will also ensure the smooth implementation of the 'Interim Measures for the Management of AI Anthropomorphic Interactive Services'."
Regarding key elements like talent, computing power, and capital, the Intelligent Agent Measures also propose specific guarantee plans. According to the latest information shared by Li Yantao, the Beijing Economic-Technological Development Area has taken the lead in releasing OPC talent identification standards. Eligible OPCs are integrated into the Yicheng talent system, supported by a social security continuation and insurance safety net during the entrepreneurial transition period, providing worry-free protection for OPC talent entrepreneurship. It has also issued the country's first virtual idol identity authentication and innovated a sandbox approval mechanism, allowing pilot enterprises to conduct digital human commercial performances within the sandbox, thereby accelerating the commercialization of new business forms.
Huge amount of information, precise interpretation, all in Sina Finance APP
Comments