近日，云蝶科技完成数亿元B+轮系列融资。本轮由StellarS Capital、亚洲联合基建（00711.HK）、长浙投资等知名机构和个人联合投资。资金将主要用于具身智能核心技术迭代与真实场景规模化落地，重点提升机器人持续工作的可靠性、跨场景部署效率及长期服务能力。云蝶科技成立于2019年，总部位于广州，是一家以大模型和智能体技术为核心的人工智能公司，业务涵盖数字智能与具身智能。在具身智能领域，云蝶科技推动机器人学习、上岗到规模化工作。公司核心团队成员来自北京大学、新加坡国立大学、南洋理工大学、香港大学等高校，能力覆盖世界预测、空间认知、多智能体决策与机器人执行。截至目前，云蝶科技已服务全国数千家政企客户。
Recently, Yundie Technology completed a Series B+ funding round worth hundreds of millions of yuan. This round was jointly invested by well-known institutions and individuals including StellarS Capital, Asia United Infrastructure (00711.HK), and Changzhe Investment. Funds will mainly be used for embodied intelligence core technology iteration and scaled real-scenario deployment, focusing on improving robots' continuous operation reliability, cross-scenario deployment efficiency, and long-term service capability. Yundie Technology was founded in 2019, headquartered in Guangzhou. It is an AI company centered on large models and agent technology, with business covering digital intelligence and embodied intelligence. In embodied intelligence, Yundie Technology advances robots from learning and onboarding to scaled work. Core team members come from Peking University, National University of Singapore, Nanyang Technological University, and the University of Hong Kong, with capabilities covering world prediction, spatial cognition, multi-agent decision-making, and robot execution. To date, Yundie Technology has served thousands of government and enterprise customers nationwide.
云蝶认为，机器人必须率先深入真实场景，利用真实业务产生的感知、动作、结果、接管闭环数据驱动能力持续进化。公司与北京大学联合提出全球首个具身环境智能（EEI）范式，构建系统能力、执行载体、行业工作流、持续服务全链路闭环，从酒店、物业、医疗康养和园区巡检等高频、可约束、可评测的B端场景切入，让真实场景既是交付现场，也是训练和验证系统。围绕真实上岗目标，云蝶现已形成RoboForge系统级具身大脑、WING任务型本体家族与行业工作流相结合的完整产品体系。RoboForge由云蝶科技和新加坡南洋理工大学联合发布，已在EmbodiedBench与LaRA两项国际权威公开基准评测中全球登顶，具备跨场景、跨任务、跨本体的迁移能力。WING任务型本体家族坚持先定义岗位、再定义机器人，打通酒店服务、医疗导诊、药房作业、物业与园区运维等典型行业工作流。
Yundie believes robots must first go deep into real scenarios, using closed-loop data generated by real business—perception, action, result, takeover—to drive continuous capability evolution. The company, together with Peking University, proposed the world's first Embodied Environment Intelligence (EEI) paradigm, building a full-chain closed loop of system capability, execution carrier, industry workflow, and continuous service, entering from high-frequency, constrainable, evaluable B2B scenarios including hotels, property management, medical care, and park inspection—making real scenarios both delivery sites and training/validation systems. Around the "real onboarding" goal, Yundie has formed a complete product system combining the RoboForge system-level embodied brain, the WING task-oriented body family, and industry workflows. RoboForge, jointly released by Yundie Technology and Nanyang Technological University, has topped global rankings on two authoritative public benchmarks, EmbodiedBench and LaRA, with cross-scenario, cross-task, and cross-body transfer capability. The WING task-oriented body family adheres to "define the position first, then define the robot," connecting typical industry workflows including hotel services, medical guidance, pharmacy operations, and property/park maintenance.
在商业模式上，云蝶推进RaaS持续运营模式，将机器人、场景部署、任务技能和持续运维整合为服务，以任务成功率、人工接管率、故障恢复时间和单位任务成本检验交付质量，构建可持续的数据飞轮。公司构建1+2+N真实世界具身智能联合研究计划，以1套联合机制为纽带，依托联合实验室与训练场2类载体，辐射N个真实场景节点。目前，云蝶已与北京大学、华南理工大学和大湾区大学等高校建立联合实验室，并全力推进北大×云蝶技术创新中心和广州海珠真实世界实验场建设。同时与云迹科技、领益智造、技象科技等伙伴深化产业协同。
On the business model, Yundie advances the RaaS continuous operation model, integrating robots, scenario deployment, task skills, and continuous operations into services, testing delivery quality by task success rate, manual takeover rate, fault recovery time, and unit task cost, building a sustainable data flywheel. The company built a "1+2+N" real-world embodied intelligence joint research plan: with 1 joint mechanism as the link, relying on 2 types of carriers—joint laboratories and training grounds—radiating to N real scenario nodes. Currently, Yundie has established joint laboratories with Peking University, South China University of Technology, and the Greater Bay Area University, and is fully advancing construction of the PKU×Yundie Technology Innovation Center and Guangzhou Haizhu Real World Experimental Field. It is also deepening industrial collaboration with partners including Yunji Technology, Lingyi iTech, and Jixiang Technology.
StellarS Capital创始合伙人Mao Zhang表示，看好云蝶科技以环境增强技术路线为核心，依托系统级具身大脑构建智能复利的增长路径。现场任务持续产生成功、失败与接管数据，数据用于改善系统，提升后的能力再服务更多场景。亚洲联合基建董事局主席彭一庭表示，作为投资者，不仅支持云蝶发展，更希望结合自身在建筑工程、基建施工及设施管理的场景，与云蝶合作推动具身智能在工地及日常运营中落地。云蝶科技创始人兼董事长田雪松表示，本轮融资完成后，将重点提升机器人的稳定性和部署效率，完善持续运营与服务能力，加速机器人从会展示走向能工作。
StellarS Capital founding partner Mao Zhang said he is bullish on Yundie Technology's growth path centered on environment-augmentation technology, building "intelligence compound interest" through a system-level embodied brain. On-site tasks continuously generate success, failure, and takeover data; data is used to improve the system; improved capabilities then serve more scenarios. Asia United Infrastructure Chairman Peng Yiting said that as an investor, the group not only supports Yundie's development but hopes to combine its own scenarios in construction engineering, infrastructure construction, and facility management to collaborate with Yundie in deploying embodied intelligence in construction sites and daily operations. Yundie Technology founder and chairman Tian Xuesong said after this funding round, the company will focus on improving robot stability and deployment efficiency, refining continuous operations and service capabilities, accelerating robots from "can demonstrate" to "can work."
