自变量机器人近日宣布完成新一轮数亿元融资，由上汽金控领投。叠加今年1月完成的十亿元级融资，该公司一年内连续获得两笔重磅资金，投后估值稳居国内具身智能赛道第一梯队。
AgiBot recently announced the completion of a new funding round worth hundreds of millions of yuan, led by SAIC Financial Holdings. Combined with the ten-billion-yuan-level round completed in January this year, the company has secured two major funding rounds within one year, with post-money valuation firmly in the first tier of China's embodied intelligence track.
自变量机器人专注于端到端通用具身智能基础模型研发。2026年4月，公司发布世界统一模型WALL-B，系全球首个将视觉、语言、触觉、动作、物理预测整合进同一神经网络的具身大模型。同年6月，公司开源WALL-OSS-0.5并发布世界模型WALL-WM，同时推出TwinDEX灵巧操作手与QUANXTA Zero无本体数据采集方案。
AgiBot focuses on end-to-end general embodied intelligence foundation model R&D. In April 2026, the company released the World Unified Model WALL-B, the world's first embodied foundation model integrating vision, language, touch, action, and physical prediction into the same neural network. In June of the same year, the company open-sourced WALL-OSS-0.5 and released the world model WALL-WM, while launching the TwinDEX dexterous manipulation hand and the QUANXTA Zero bodyless data collection solution.
商业化方面，公司与58到家合作推出机器人上门家政服务，已从深圳拓展至北京，2026年7月入选APEC《亚太地区数智赋能案例集》。工业场景下，物流分拣方案效率达1816件/小时、准确率超98%，在2026世界机器人大会上实测提升至1911件/小时。公司创始人兼CEO王潜毕业于清华大学，是全球最早提出注意力机制的研究人员之一；联合创始人兼CTO王昊为北京大学计算物理博士，曾主导发布多模态开源大模型"太乙"及百亿级大语言模型"燃灯"、千亿级"姜子牙"。
On commercialization, the company partnered with 58到家 to launch robot home housekeeping services, expanding from Shenzhen to Beijing, and was selected for APEC's "Asia-Pacific Digital Intelligence Empowerment Case Collection" in July 2026. In industrial scenarios, its logistics sorting solution achieves 1,816 items per hour with over 98% accuracy, measured at 1,911 items per hour at the 2026 World Robot Conference. Founder and CEO Wang Qian graduated from Tsinghua University and is among the world's earliest researchers to propose attention mechanisms; co-founder and CTO Wang Hao holds a Ph.D. in computational physics from Peking University and previously led the release of the multimodal open-source model "Taiyi," the tens-of-billions-parameter large language model "Randeng," and the hundred-billions-parameter "Jiangziya."
这轮融资的领投方是上汽金控，这个信息值得单独看。上汽金控是上汽集团的产业投资平台，它投自变量机器人，逻辑和前面墨现科技那轮里自变量机器人投资触觉传感器公司是一样的：产业资本通过投资上游技术公司，锁定未来可能用到的能力。对上汽来说，具身智能大模型如果成熟，汽车制造和物流环节的分拣、搬运、装配都有可能用上。对自变量机器人来说，拿到上汽金控的钱，不只是拿到资金，还拿到了一个潜在的工业场景入口。前面提到的物流分拣方案效率数据，1816件/小时、准确率超98%，这个性能指标如果能在上汽的工厂里跑通，商业化的路径会更清晰。
The lead investor in this round is SAIC Financial Holdings, a piece of information worth examining separately. SAIC Financial Holdings is SAIC Group's industrial investment platform. Its investment in AgiBot follows the same logic as AgiBot's investment in the tactile sensor company in the earlier Moxian Technology round: industrial capital locks in capabilities it may need in the future by investing in upstream technology companies. For SAIC, if the embodied intelligence foundation model matures, sorting, handling, and assembly in automotive manufacturing and logistics could all use it. For AgiBot, taking SAIC Financial Holdings' money means not just capital, but a potential industrial scenario entry point. The logistics sorting efficiency data mentioned earlier—1,816 items per hour with over 98% accuracy—if this performance metric can run in SAIC's factories, the commercialization path becomes clearer.
自变量机器人一年内两轮融资，投后估值稳居第一梯队，这个节奏说明具身智能基础模型这条赛道，资本仍在集中押注头部。前面讨论过宇树科技的股价波动、云深处的技术复用路径、多家数据采集公司的商业模式，自变量机器人走的是另一条路：不做硬件本体，不做数据采集服务，做的是整合视觉、语言、触觉、动作、物理预测的端到端基础模型。这条路的技术门槛最高，烧钱也最快，但一旦跑通，它能同时为家政、物流、工业多个场景提供底层能力。上汽金控这轮投资，是对这条路线的又一次确认。
AgiBot's two funding rounds within one year, with post-money valuation firmly in the first tier, indicates that in the embodied intelligence foundation model track, capital is still concentrating on leading players. Earlier we discussed Unitree's stock price fluctuations, Deep Robotics' technology reuse path, and the business models of multiple data collection companies. AgiBot takes a different path: not making hardware bodies, not doing data collection services, but building an end-to-end foundation model integrating vision, language, touch, action, and physical prediction. This path has the highest technical threshold and burns cash fastest, but once it works, it can provide underlying capabilities for housekeeping, logistics, and industrial scenarios simultaneously. SAIC Financial Holdings' investment this round is another confirmation of this route.
