CtrlK
| Robbyant R1 | |
|---|---|
| General information | |
| Manufacturer | Robbyant (Ant Lingbo Technology), subsidiary of Ant Group |
| Country of origin | China |
| Year unveiled | 2025 |
| Status | In production |
| Type | Wheeled humanoid robot |
| Estimated price | ~$55,000 USD (bundled as scenario solutions) |
| Website | robbyant.com |
The Robbyant R1 is a wheeled humanoid robot developed by Robbyant (formally known as Ant Lingbo Technology, or Shanghai Mayi Lingbo Technology), a robotics subsidiary of Ant Group, the Chinese fintech giant behind Alipay. Unveiled in September 2025 at both the IFA consumer electronics trade show in Berlin and the Inclusion Conference on the Bund in Shanghai, the R1 is a dual-arm, wheeled humanoid designed for service applications in hospitality, catering, healthcare, and public venues. Powered by Ant Group's proprietary large language models and embodied AI technology, the R1 represents the first hardware product from Jack Ma's fintech empire as it expands into the rapidly growing humanoid robotics sector.
The R1 attracted international attention for its live cooking demonstrations, during which it autonomously prepared garlic shrimp and other dishes for audiences, handling ingredient selection, wok operation, stirring, and cleanup without human intervention. As of late 2025, the robot had entered mass production and been delivered to clients including the Shanghai History Museum. In early 2026, the Robbyant R1 received an iF Design Award in the Product Design category.
Ant Group, valued as one of the world's largest fintech companies, operates the Alipay digital payments platform serving over one billion users. Founded by Jack Ma as an affiliate of Alibaba Group, Ant Group has historically focused on financial technology, digital payments, and cloud services. The company's move into robotics reflects a broader strategic shift among major Chinese technology companies toward embodied intelligence, the application of artificial intelligence to physical systems that can perceive and interact with the real world through sensors and actuators.
In December 2024, Jack Ma stated that the company "must embrace the fast-developing technology" and that "changes brought by AI in the next 20 years will go beyond everyone's imagination." [1] Shortly after, Ant Group established Shanghai Mayi Lingbo Technology (later branded as Robbyant) in December 2024, registering the company with RMB 100 million ($13.7 million) in capital. [2] The subsidiary formally launched in Shanghai's Pudong District in March 2025, with Ant Group chair Eric Jing Xiandong and chief executive Cyril Han Xinyi in attendance. [3] A second facility opened in Hangzhou in August 2025 to expand research and development operations. [3]
Robbyant is led by CEO Zhu Xing, who has articulated the company's mission as leveraging "embodied intelligence to extend the services Ant provides in the digital world more effectively into the physical world." [3] From early 2025, the company aggressively recruited robotics talent through the Chinese platform Boss Zhipin, offering annual salaries of up to RMB 1.04 million ($142,800) for humanoid robot motion control engineers and up to RMB 1.12 million ($153,800) for hardware structure designers. [2] The job descriptions covered designing and optimizing robotic arm and human motion control systems, integrating algorithms with software and hardware components, and creating detailed design specifications for robot construction.
Ant Group's entry into robotics placed it alongside other major Chinese technology firms, including Baidu, Tencent, Huawei, JD.com, Xiaomi, and Xpeng, all of which have invested in robotics and AI-driven physical systems.
The Robbyant R1 is a wheeled, dual-arm humanoid robot with a white and metallic finish. Rather than using bipedal locomotion, the R1 relies on a wheeled base for stability and mobility, prioritizing manipulation dexterity over walking capability. The robot's upper body features a humanoid torso with two articulated arms designed for complex object manipulation tasks.
| Category | Specification | Details |
|---|---|---|
| Dimensions | Height | 1,600 to 1,750 mm (adjustable) |
| Dimensions | Weight (with battery) | 110 kg (243 lbs) |
| Mobility | Locomotion type | Wheeled base |
| Mobility | Maximum speed | 1.5 m/s (5.4 km/h, 3.4 mph) |
| Manipulation | Degrees of freedom | 34 (primarily in the arms) |
| Manipulation | Arm configuration | Dual-arm, dexterous manipulation |
| Actuators | Motor type | 48V DC geared motors with encoders |
| Actuators | Joint actuators | In-house developed modular joint actuators |
| Perception | Vision system | Multimodal perception cameras |
| Perception | Capabilities | Object recognition, spatial mapping, environmental understanding |
| AI | Primary LLM | Bailing (Ling series) large language model |
| AI | Architecture | Cloud-assisted processing with edge inference |
| AI | Task planning | End-to-end, long-horizon autonomous planning |
| Connectivity | Wireless | Bluetooth, Wi-Fi |
| Connectivity | Cloud integration | Real-time AI processing and updates |
| Software | Platform | Robbyant Open Platform |
| Software | License | Closed source (robot); open-source AI models |
The 34 degrees of freedom are concentrated primarily in the robot's arms, enabling dexterous tasks such as grasping ingredients, operating kitchen equipment, and manipulating utensils. Some sources have reported up to 43 degrees of freedom; the discrepancy may reflect different configurations or the inclusion of additional end-effector articulation. [4][5] The robot's 48V DC geared motors with encoders provide precise positional control for each joint, while the in-house developed modular joint actuators allow for customization across different deployment scenarios.
The R1 uses a multimodal perception system that combines visual cameras with spatial awareness capabilities. The vision system enables the robot to recognize objects, map its environment in three dimensions, and understand spatial relationships between items such as tables, appliances, cookware, and utensils. This spatial perception is critical for the robot's core kitchen and service tasks, where it must identify the positions of ingredients and cooking equipment in real time.
In January 2026, Robbyant partnered with Orbbec, a 3D sensing technology company, to develop and deploy LingBot-Depth, a high-precision spatial perception AI model. Orbbec's Gemini 330 stereo 3D camera, which uses a proprietary MX6800 depth engine chip, was selected for integration. The partnership aims to incorporate LingBot-Depth into next-generation depth cameras specifically designed for embodied intelligence applications. [6]
The R1 is powered by Ant Group's proprietary Bailing large language model, which forms part of the broader Ling series of AI models. The Ling series includes models at different scales: Ling-Lite with 16.8 billion parameters (2.75 billion activated) and Ling-Plus with 290 billion parameters (28.8 billion activated), both using a mixture-of-experts (MoE) architecture. [7] Notably, Ant Group published research in March 2025 demonstrating how to efficiently train a 300-billion-parameter MoE model on domestic (non-premium) GPUs, reducing reliance on restricted high-end chips. [7]
The AI system enables the R1 to perform end-to-end, long-horizon task planning. Rather than following simple pre-programmed routines, the robot can autonomously decompose complex tasks into sequential steps, adapt to changing conditions, and continuously learn from new experiences. This architecture allows the robot to plan and execute workflows such as receiving a meal order, selecting appropriate ingredients, operating cooking equipment, preparing the dish, and delivering the finished product.
The robot employs a dual-driven embodied control algorithm powered by both simulation-based training and real-world data. According to the iF Design Award description, this approach achieves "millisecond-level responsiveness and action efficiency close to that of humans." [8]
The R1's signature demonstration has been autonomous cooking. At its unveiling at the 2025 Inclusion Conference on the Bund in Shanghai (September 11, 2025) and at IFA 2025 in Berlin (September 5 to 9, 2025), the robot performed live cooking demonstrations that drew significant media attention. During the Shanghai event, the R1 prepared four complete dishes for the audience, demonstrating its ability to recognize ingredients, locate and operate utensils, work a wok and stove, stir food, serve completed dishes, and clean up after itself. [9]
At the IFA 2025 booth, one R1 unit was placed on static display while a second was deployed in a mock kitchen environment, actively cooking garlic shrimp for visitors. The robot selected ingredients, placed them into a wok, operated the stove, and completed the dish autonomously. [10]
Robbyant has claimed that the R1 has access to a digital cookbook containing over 10,000 dishes and 1,000 tea drink recipes, with each dish reportedly taking approximately 10 minutes to prepare. [9] The system is compatible with different brands of woks, stoves, and coffee machines, and can be customized for specific kitchen layouts and workflows. [11]
Beyond cooking, the R1 has been deployed as an automated tour guide. The Shanghai History Museum became one of the robot's earliest clients, using R1 units to provide guided tours for visitors. The robot's multimodal perception system, combined with its natural language processing capabilities powered by the Bailing LLM, allows it to navigate museum spaces, identify exhibits, and provide spoken explanations to visitors. [3]
Robbyant has also positioned the R1 for healthcare applications, including sorting medicine in pharmacies and providing basic medical consultations. The robot's dexterous manipulation capabilities make it suitable for handling pharmaceutical products, while its AI-powered conversational abilities enable basic patient interaction. [3][10]
Through the Robbyant Open Platform, ecosystem partners can deploy the R1 across various daily-life service venues including supermarkets, exhibition halls, and offices. Use cases include marketing and shopping guidance, guided tours, and food preparation. [8] The company has indicated that future development will expand into robotic companions and caregivers for consumer markets and underserved sectors such as elderly care. [11]
Rather than selling the R1 as a standalone hardware product, Robbyant bundles the robot into broader "scenario solutions" for businesses and institutions. [3] This approach reflects Ant Group's fintech DNA, where platform-based services generate recurring revenue rather than one-time hardware sales. The model allows Robbyant to provide end-to-end deployment support, including customization for specific environments, ongoing software updates through cloud connectivity, and integration with the customer's existing operational workflows.
The estimated unit price of the R1 is approximately $55,000 USD, though exact pricing varies depending on the deployment scenario and bundled services. [4] Target markets include hospitality and catering operations, museums and exhibition venues, healthcare facilities and pharmacies, community centers and elderly care institutions, and commercial retail environments.
As of late 2025, Robbyant confirmed that the R1 had entered mass production. The first-generation robot was delivered to multiple early clients:
| Client | Application | Status |
|---|---|---|
| Shanghai History Museum | Tour guiding | Deployed |
| Community care centers | Service assistance | Pilot testing |
| Commercial restaurants | Cooking and food preparation | Pilot testing |
At IFA 2025 in Berlin, Robbyant's DACH Country Manager Zheng Yuewen announced the company's intention to seek European partners for practical applications. [11] The company has framed the R1's expansion into Europe as part of a strategy to address labor shortages in hospitality, healthcare, and domestic service industries across developed economies.
In January 2026, Robbyant launched its "Evolution of Embodied AI Week" initiative, releasing a series of open-source AI models designed to accelerate the broader robotics industry. These releases positioned Robbyant not only as a robot manufacturer but also as an AI platform provider for the embodied intelligence ecosystem.
LingBot-VLA is a vision-language-action (VLA) model designed as a "universal brain" for real-world robotics. [12] The model was pre-trained on more than 20,000 hours of real-world interaction data, a dataset roughly equivalent in scale to what U.S. startup Physical Intelligence used for its Pi-0.6 model. [13] Key features include:
CEO Zhu Xing stated that LingBot-VLA is "Ant Group's first open-source embodied AI model" and represents "another milestone in their efforts toward Artificial General Intelligence (AGI)." [12]
LingBot-Depth is a high-precision spatial perception model developed in partnership with Orbbec. [6] The model uses a technique called Masked Depth Modeling (MDM) to address the challenge of depth perception in complex environments. When depth sensor inputs are incomplete or corrupted, the system analyzes RGB image features (including texture, object contours, and scene context) to reconstruct missing regions and produce denser, more accurate 3D depth maps. [6]
In benchmark evaluations on the NYUv2 and ETH3D datasets, LingBot-Depth outperformed competing models such as PromptDA and PriorDA, achieving a relative error (REL) reduction of over 70% in indoor scenes and reducing root mean square error (RMSE) by approximately 47% on sparse Structure-from-Motion tasks. [6]
LingBot-World is a world model that became the first state-of-the-art-level world model to be fully open-sourced. [14] Built on top of Wan2.2, a 14-billion-parameter image-to-video diffusion transformer, LingBot-World extends the architecture into a mixture-of-experts diffusion transformer with two experts of approximately 14 billion parameters each (28 billion total), with only one expert active at each denoising step. [14] Notable capabilities include:
The model is designed for applications in embodied intelligence training, autonomous driving simulation, and game development. It was released under the Apache 2.0 License. [14]
In March 2026, Robbyant announced a strategic partnership with Leju Robot, a Chinese robotics company specializing in humanoid robot hardware. [15] The collaboration focuses on combining Leju's strengths in robotic embodiment, data collection, and real-world use cases with Robbyant's expertise in embodied AI models. Previously, Leju had served as a core data partner, providing nearly 10,000 hours of high-quality, multimodal real-robot interaction data used to train the LingBot-VLA model. [15] The expanded partnership aims to accelerate the transition of embodied AI robots from task-specific execution to general intelligence.
The partnership with Orbbec, announced in January 2026, centers on integrating LingBot-Depth into Orbbec's next-generation depth cameras. Orbbec's Gemini 330 stereo 3D camera, equipped with the company's proprietary MX6800 depth engine chip, was specifically developed for robotics applications and delivers reliable 3D data across extreme lighting conditions, from complete darkness to direct sunlight. [6]
The Robbyant R1 entered a rapidly expanding Chinese humanoid robot market. By 2025, Chinese companies controlled an estimated 90% of the global humanoid robot market by volume. [16] Between 13,000 and 18,000 humanoid robots were sold worldwide in 2025, with Chinese manufacturers leading in production and deployment. [16]
Key competitors include:
| Company | Robot | Type | Notable facts |
|---|---|---|---|
| Unitree Robotics | Unitree H1, G1 | Bipedal humanoid | Sold 5,500 units in 2025; filed for Shanghai IPO |
| AgiBot | Multiple models | Bipedal humanoid | Shipped 5,168 units in 2025; reached 10,000 by March 2026 |
| Tesla | Optimus | Bipedal humanoid | Target of 5,000 units in 2025 (not met); consumer sales expected by 2027 |
| Figure AI | Figure 02 | Bipedal humanoid | Backed by major U.S. investors; advanced AI integration |
| UBTech | Walker S2 | Bipedal humanoid | Chinese leader in bipedal humanoid development |
| Xpeng Robotics | Iron | Bipedal humanoid | Backed by EV manufacturer Xpeng |
The R1 differentiates itself from most competitors through its wheeled rather than bipedal design, its focus on manipulation dexterity over locomotion, and its bundled scenario-solution business model. While companies like Unitree have driven prices as low as $5,900 for basic bipedal humanoids, the R1 targets a different market segment with its emphasis on commercial service deployments and integrated AI capabilities. [16]
As Unitree Robotics founder Wang Xingxing has observed, "The hardware for robot bodies is already sufficient. The real bottleneck lies in embodied intelligence models, whose capabilities are still far from adequate." [3] Robbyant's strategy of open-sourcing its AI models while selling integrated solutions positions it to address this intelligence bottleneck across the industry.
China's humanoid robot market was projected to reach RMB 12 billion by 2030 with 1.5 million unit shipments, while global estimates from RBC Capital Markets suggest the humanoid robot market could reach $9 trillion by 2050, with China accounting for over 60% of the total. [2][16]
Robbyant has confirmed that a second-generation model is under development. [3] The company has indicated that when fully developed, future versions of the R1 platform may transition from a wheeled base to legged locomotion. [9] The second-generation robot is expected to incorporate improvements in manipulation dexterity, AI capabilities, and deployment flexibility.
Robbyant CEO Zhu Xing has stated that the company's overarching goal is "to accelerate the integration of AI into the physical world so it can deliver practical value sooner." [13] The company's dual strategy of producing its own hardware while open-sourcing foundational AI models positions Robbyant as both a robotics manufacturer and an ecosystem enabler for the broader embodied intelligence industry.