On June 12, 2026, the 8th BAAI Conference officially commenced at the Zhongguancun (ZGC) International Innovation Center. Hosted by the Beijing Academy of Artificial Intelligence (BAAI), a world-leading non-profit AI research institute, this conference has evolved into one of the world’s premier academic conferences in artificial intelligence. Every June, the world’s leading academics gather in Beijing to discuss the latest advances in technology. This year’s conference achieved an unprecedented scale, drawing more than 10,000 on-site participants and over 5 million online viewers. As of its 8th edition, the BAAI Conference has hosted 14 Turing Award laureates. Sam Altman is also among the distinguished guests who have delivered keynote addresses at the summit.
It is widely recognized for its emphasis on cutting-edge technologies, global vision and the cultivation of emerging talents. The conference convenes researchers, industry executives and young scientists across the globe to present The latest research breakthroughs, delve into cutting-edge theories academic theories and share industry insights.
This year's conference features an unprecedented gathering of leading AI researchers. Whitfield Diffie, Turing Award laureate and pioneer of modern public-key cryptography, attended in person to discuss security and trust challenges in the emerging Agent era. Andrew Barto, Turing Award laureate and one of the pioneers of reinforcement learning, explored the implications of interaction-driven intelligence for the next generation of AI systems.
More than 200 leading scholars, over 40 AI company CEOs, founders and chief scientists, and more than 30 young scientists under the age of 30 have gathered in Beijing. For the first time, many of China's most influential innovators in world models and AI agents are appearing together on a single stage.
The conference also brings together thousands of prominent researchers from across the global AI community to engage in discussions on world models, general-purpose AI agents, embodied intelligence, AI safety, AI-native education, the token economy and OPC, as well as next-generation intelligent computing architectures.
The opening ceremony was chaired by Huang Tiejun, Chairman of the BAAI Board of Directors.
Wang Zhongyuan, President of BAAI, delivered the institute's "2026 Research Progress Report," presenting the latest advances in foundation models, AI agents, foundational software and hardware ecosystems, as well as updates on BAAI's open-source initiatives.
Since its establishment in 2018, BAAI has released the WuDao series of foundation models and the WuJie family of world models, building a comprehensive open-source technology ecosystem for large AI models from the ground up. From the early stages of foundation model development to today's emerging era of embodied AI, BAAI has remained at the forefront of AI research and innovation.
To date, BAAI has open-sourced more than 200 AI models, with cumulative global downloads exceeding one billion. The institute has also incubated a number of startups in both large-model and embodied AI sectors.
At the 2024 BAAI Conference, the institute outlined its vision for the evolution of artificial intelligence, particularly foundation models. Today, AI is rapidly advancing from large language models to multimodal models and further toward world models, accelerating the transition from the digital world into the physical world.
Over the past year, BAAI has achieved significant research breakthroughs in foundation models, AI agents, and foundational software and hardware ecosystems. Drawing on its exploration of multimodal and world models, the institute presented a systematic review of the development of world-model technologies, and introduced a four-category framework for understanding existing world-model architectures. Research on WuJie·Emu3, a native multimodal world model, was published in Nature in January 2026. BAAI also released an AI-ready neuroscience dataset and the data platform BrainToken alongside WuJie·Brainμ 1.0. WuJie·Brainμ stands as the world’s first unified multimodal neuroscience large model capable of synchronous comprehension and generation, constructing a universal multimodal neuroscience foundational base grounded in world-model technologies. The joint technical achievement, co-developed by BAAI and Tsinghua University, was published in Science.
During the opening ceremony, Whitfield Diffie and Andrew Barto each delivered keynote speeches addressing critical challenges and opportunities in the future of AI.
A featured live podcast titled "A Decade Ahead of the Curve" brought together Huang Tiejun and Wang Jian, Founder of Alibaba Cloud and Director of Zhijiang Lab. The panel explored the evolution of artificial intelligence paradigms and dissected how foundation models elevate systemic architecture development capabilities and advance paradigm innovation. It also addressed key challenges, including limitations in internet-scale text data.
As AI enters the agent era, the speakers discussed the profound opportunities and risks associated with large-scale deployment of autonomous intelligent agents in the real world. They also shared long-term perspectives on the evolving relationship between humans and AI, and on how human civilization and AI systems may coexist in the future.
In the roundtable session titled "Reshape the World: A Summit Dialogue on Foundation Models," Wang Zhongyuan joined leading AI researchers and industry pioneers, including Zhu Jun, Professor at the Department of Computer Science and Technology of Tsinghua University; Luo Fuli, Head of Xiaomi's MiMo Project; Liu Zhiyuan, Professor at Tsinghua University; and Bo An, Presidential Chair Professor and Dean of the College of Computing and Data Science at Nanyang Technological University.
The panel explored the evolution of advanced foundation model capabilities, AI RSI (recursive self improvement), multimodal intelligence, and world models, examining how AI is moving beyond digital environments and increasingly interacting with the physical world.
Participants agreed that increasingly capable foundation models and agent systems are opening new possibilities, while self-evolving AI may become a catalyst for future advances in AI. The rapid maturation of multimodal and world-model technologies is expected to accelerate AI's transition from digital spaces into real-world applications. They also emphasized the importance of creating broader opportunities for young researchers to participate in shaping the future of AI.
BAAI continues to strengthen the foundations of global AI innovation through academic breakthroughs, ecosystem development, startup incubation, and comprehensive open-source initiatives.
Media Contact
Yang Xiyang
xyyang01@baai.ac.cn