Digital News

　　The 2025 Artificial Intelligence (AI) Computing Conference, themed “Laying Foundations, Embracing Openness, and Scaling Globally,” took place on September 26 in Beijing. The conference delved into cutting-edge topics such as open computing, the low-altitude economy, AI for Science, intelligent computing centers, and autonomous driving. It aimed to fully implement the national strategy for accelerating the development of new quality productive forces and actively respond to the State Council’s plan for the in-depth rollout of the “AI Plus” initiative. With the core element of computing power as the driver of innovation, the conference brought together forces from the industry, universities, research institutions, and end-users to collectively advance the high-quality development of the AI industry.

　　The conference was jointly guided by the Beijing Municipal Commission of Development and Reform, Beijing Municipal Science & Technology Commission, Administrative Commission of Zhongguancun Science Park, Beijing Municipal Bureau of Economy and Information Technology, and the Haidian District People’s Government, with support from the Chinese Association for Artificial Intelligence (CAAI). It was co-hosted by Tsinghua University, Beijing Academy of Artificial Intelligence (BAAI), Zhongguancun Academy, Zhongguancun Institute of Artificial Intelligence, OCP Foundation, Open Compute Technology Committee (OCTC) of the China Electronics Standardization Association, along with leading enterprises in the AI industry. The conference unveiled four key innovative outcomes, creating a closed loop that spans computing power supply, technological collaboration, standards support, and independent innovation, thereby laying a strong foundation for the industrialization and large-scale implementation of the “AI Plus” initiative.

　　“Beijing Solution” for Super-node Intelligent Computing Applications Builds the Core of Inclusive Computing Power Supply

　　Computing power, as the “engine” driving the AI development, is essential for the large-scale implementation of the “AI Plus” initiative. The conference unveiled the “Beijing Solution” for Super-node Intelligent Computing Applications (hereinafter referred to as the Beijing Solution), a major milestone in building an integrated and inclusive computing power foundation for the “AI Plus” initiative.

　　The Beijing Solution is centered on the core principles of “openness and diversity, full-scenario empowerment, and universal accessibility.” It integrates key elements such as chips, complete systems, models, services, and applications to build a fully domestic industrial chain, fostering a virtuous cycle of “technological innovation—AI industry upgrade—intelligent transformation of society and the economy.”

　　The Beijing Solution was initiated by the Beijing Municipal Science & Technology Commission, Administrative Commission of Zhongguancun Science Park, with participation from complete system manufacturers, AI chip enterprises, large model developers, AI cloud service providers, industrial software developers, system integrators, and research institutions, including BAAI and Tsinghua University. The Solution focuses on more than ten key scenarios, including AI for Science, embodied intelligence, intelligent manufacturing, and smart healthcare. A signing ceremony for the first batch of users was held concurrently during the conference.

　　Full-Stack Open-Source Technology PlatformFlagOS 1.5 Overcomes Ecological Collaboration Challenges

　　To integrate the entire industrial chain, it is urgent to address key industry challenges, including the fragmentation of the AI ecosystem and the difficulty in migrating software and hardware. Open-source and openness are essential for breaking down technological barriers and fostering collaboration across the industry. To this end, with support from the Beijing Municipal Commission of Development and Reform, the BAAI has worked with global ecosystem partners to officially launch the “FlagOS 1.5: Large Model Full-Stack Open-Source Technology Platform.”

　　BAAI, in collaboration with multiple manufacturers and institutions, has spearheaded the development of FlagOS, a unified open-source system software technology stack tailored for diverse AI chips. This stack includes essential components such as the efficient parallel training and inference framework FlagScale, the high-performance operator library FlagGems, and the unified communication library FlagCX. Through its open-source approach, FlagOS fosters underlying technology collaboration and innovation, providing a decoupled, efficient, and user-friendly development environment for “AI Plus,” while accelerating the adoption and integration of technology into applications.

　　Lin Yonghua, Vice President and Chief Engineer of BAAI, referred to FlagOS as the “operating system for large AI models.” Its main goal is to provide unified, open, and efficient computing support for AI models, particularly in heterogeneous computing power environments. FlagOS tackles compatibility issues between different AI chips and addresses the fragmentation of the AI hardware ecosystem, enabling large models to run efficiently across diverse hardware platforms. This will lay the groundwork for a globally unified, open-source, full-stack AI ecosystem.

　　Beijing Artificial Intelligence Standardization Technical Committee Strengthens Compliance Framework

　　The implementation of the computing power foundation and the expansion of the open-source ecosystem urgently require unified standards, as these are essential for ensuring the efficient coordination, security and controllability of computing resources. In this context, the Beijing Artificial Intelligence Standardization Technical Committee (hereinafter referred to as the “Standardization Technical Committee”) was officially established at the conference.

　　The Standardization Technical Committee will provide professional technical support for project approval, drafting, initial review, implementation evaluation, and revision of local standards in the AI field in Beijing. By developing and promoting standards, it will not only ensure the steady and secure progress of “AI Plus” but also facilitate the standardized application of technological achievements, thus removing barriers to cross-industry and cross-sector collaboration.

　　Reconfigurable Computing Power Hardware-Software Collaborative Innovation Center Strengthens the Independent Foundation

　　The collaborative development of computing power, ecosystems, and standards depends on a secure, controllable, and efficiently adaptable computing power foundation. To strengthen the domestic foundation for “AI Plus,” TSING MICRO, in collaboration with industry partners and with support from the Beijing Municipal Government, has established the “Beijing Reconfigurable Computing Power Hardware-Software Collaborative Innovation Center,” along with the launch of an innovation consortium.

　　The center brings together BAAI, Tsinghua University, Business-Intelligence of Oriental Nations Corporation (BONC), and other organizations, aiming to collaboratively optimize reconfigurable computing chip architectures and software ecosystems. It aims to introduce new AI solutions for intelligent computing centers, vertical industries, and the internet, establishing an independent and controllable computing power foundation for the high-quality development of AI. By fostering innovative technology chains, the center will accelerate the digital and intelligent upgrading of the industry chain.

　　The conference also featured a 3,000-square-meter immersive AI Innovation Technology Exhibition, showcasing major innovations and practical applications in AI. Key highlights included the super-node achievements related to the Beijing Solution and the world’s first passenger-carrying unmanned electric vertical takeoff and landing (eVTOL) aircraft to receive model certification.