The online preliminary round of the first Ant Lingbo Embodied Large Model Challenge concluded: 1,855 teams registered

The online preliminary round of the first "Ant Lingbo Embodied Large Model Challenge" has come to a close. According to the organizers, the competition attracted a total of 1,855 teams and 3,252 contestants from home and abroad. Participants came from 13 countries, covering 648 well-known universities, research institutes, leading technology companies and embodied intelligence companies, including 41 overseas universities and organizations, making the inaugural competition the largest of its kind in the industry. In September this year, Ant Lingbo Technology, together with ModelScope Community and Alibaba Cloud Tianchi, launched the first "Ant Lingbo Embodied Large Model Challenge" at the 2026 Bund Conference. This is an embodied intelligence algorithm challenge that uses the open-source model LingBot-VLA2.0 as its technical foundation, focusing on model training and tuning, simulation generalization evaluation and real-robot adaptation, and is open to enterprises, universities, individual developers and technology enthusiasts worldwide. At 0:00 on October 27, registration for the online preliminary round closed. A total of 1,855 teams signed up for the competition, attracting students from prestigious universities such as Tsinghua University, Peking University, Shanghai Jiao Tong University, Zhejiang University, University of Science and Technology of China, Tongji University, University of Chinese Academy of Sciences and Harbin Institute of Technology, as well as the University of Hong Kong, Hong Kong University of Science and Technology and the Chinese University of Hong Kong. Research teams from internationally renowned universities such as Carnegie Mellon University, Georgia Institute of Technology, University of Southern California, University College London, Technical University of Munich, National University of Singapore, Nanyang Technological University, University of Sydney, KTH Royal Institute of Technology and Waseda University also took part. Industry practitioners also used this competition as a technical "training ground." Developers from leading technology companies such as Huawei, Tencent, Alibaba, ByteDance, Meituan, Microsoft, BYD and Xiaomi formed teams to compete; domestic humanoid robot and embodied intelligence companies such as UBTECH, Fourier Intelligence, AgiBot, Zhuji Dynamics and the Beijing Humanoid Robot Innovation Center also all entered the arena, making this challenge a real battlefield combining frontier academic exploration with industrial engineering deployment. According to the profile released by the organizing committee, among all participants, university students and researchers accounted for more than 70% (74.7%), with master's and doctoral students making up as much as 67.4%; at the same time, developers from enterprises and frontline practical fields accounted for 25.2%, making high educational attainment and strong practical experience the distinctive labels of this year's contestants. From the perspective of the technology map, these developers are highly concentrated in the most core technical chain of embodied intelligence. Among them, contestants with backgrounds in VLA (vision-language-action) or multimodal large models accounted for as much as 51.7%; deep learning training accounted for 48%; reinforcement learning (31.9%), simulation environments (25.4%), model deployment (23.4%), real-robot debugging (20.9%) and robot control (20.1%) followed closely behind. Industry analysts pointed out that the first Ant Lingbo Embodied Large Model Challenge being able to attract more than 1,800 high-level teams from around the world in a short period of time confirms, on the one hand, that the embodied intelligence track is ushering in an unprecedented period of technological explosion, with developers eager for knowledge; on the other hand, it also demonstrates the developer community's high recognition of Ant Lingbo's self-developed embodied large models. As of now, Ant Lingbo's open-source models have accumulated more than 33,000 stars on GitHub, firmly ranking among the top domestic embodied models. Mainstream overseas technical communities have also achieved adaptation to Lingbo models. LingBot-VLA and LingBot-VA have both completed ecosystem adaptation to LeRobot, the world's largest robot development platform, and LingBot-Vision has been merged into the mainline of timm, the industry's mainstream vision backbone model library. LingBot-World has become the world's first world model to be supported by both of the two mainstream open-source inference engines, vLLM and SGLang. By opening up frontier foundation model capabilities, real physical simulation and engineering environments, Ant Lingbo not only provides a competitive stage for global developers, but also accelerates the process of the embodied intelligence "general brain" moving from theoretical algorithms to full-scenario industrial empowerment. According to the competition schedule, the more challenging offline finals and the first Ant Lingbo Developer Conference will be held in Shanghai from November 14 to 16, where 15 teams will compete in a 48-hour real-robot hackathon.