Fangqi Technology (Beijing) Co., Ltd. has officially announced the completion of a multi-million-yuan angel funding round, with participation from investors including TusStar Ventures and other institutions. The newly raised capital will be directed primarily toward advancing core technology development and expanding its R&D workforce, while also accelerating the validation and commercial deployment of key technologies within representative application scenarios.
The company intends to use these funds to continuously refine its "general-purpose skill operating system" that supports sustainable robot learning and cross-embodiment migration. This initiative is part of a broader strategy to build the foundational infrastructure for a universal "robot brain," gradually establishing a capability pipeline that enables learning, storage, and distribution of embodied intelligence across multiple robotic platforms and diverse operational environments.
According to founder Wang Xinzhou, the fundamental limitation in current embodied robot generalization stems from a disconnect: world models lack human knowledge, while the embodied "brain" is deficient in physical common sense. In simpler terms, there exists a gap between physical knowledge and semantic understanding within the world model and embodied model frameworks. To address this challenge, Fangqi Technology has introduced a semantic world model designed to create robots that can learn, work, and grow in ways that mirror human capabilities.
Founded in 2026, Fangqi Technology specializes in developing general-purpose brains for embodied intelligence. The core team members are all graduates of Tsinghua University, bringing comprehensive expertise spanning world models, VLA algorithms, cloud-native systems, and full-chain industrial commercialization. This combination provides the company with both cutting-edge research credentials and hands-on experience in real-world deployment scenarios.
The company has pioneered what it calls the "Turing Learning Paradigm," a structured approach that sequences curriculum learning, task learning, practical application, and reflective learning. This methodology breaks away from traditional teleoperation-based learning approaches that merely teach robots to replicate actions without understanding underlying principles. Instead, it enables robots to continuously learn from human data and real-world practice, establishing a scalable and repeatable learning framework.
Fangqi Technology has already validated its Turing Learning methodology and semantic world model across commercial service applications, filing multiple technology patents in the process. The company achieved second-place ranking worldwide in the Stanford BEHAVIOR 2026 household challenge competition. Strategic partnerships have been established with Youdi Robotics and Jiangsu Longhuan to advance pilot deployments, with the company positioning "three-dimensional cleaning" as its initial proving ground before expanding into broader commercial service scenarios.
TusStar Ventures General Manager and Managing Partner Liu Bo commented on the investment, noting that the competitive edge in embodied intelligence lies in moving beyond the conventional approach of massive real-world data accumulation toward finding cost-effective, scalable learning pathways for general-purpose robots. Liu expressed confidence that Fangqi Technology's semantic world model and Turing Learning paradigm offer fresh solutions to industry challenges, and that the company has strong potential to develop the next-generation foundational software platform for embodied intelligence.