The morning of April 19, 2025, dawned over Beijing’s Nanhai Zi Park with an air of the extraordinary.1 Alongside over 9,000 human runners stretching and adjusting their gear, another group of competitors prepared for the 21.0975-kilometer (roughly 13.1 miles) half marathon.3 These participants, however, were forged not of flesh and bone, but of metal, wires, and sophisticated algorithms. Twenty teams of humanoid robots, ranging from imposing 1.8-meter figures to diminutive 75-centimeter machines, stood poised on a separate, parallel track, ready to make history.1 The air crackled with a unique blend of athletic anticipation and technological curiosity, as spectators and human athletes alike craned their necks, phones aloft, capturing images of their robotic counterparts.5 Some runners quipped they weren’t aiming for a personal best (“PB”) today; they just wanted to watch the robots.17
This event, officially titled the “2025 Beijing Yizhuang Half Marathon” (人形机器人半程马拉松), marked the world’s first instance of humans and humanoid robots competing side-by-side in a sanctioned half marathon.1 It was a moment where decades of science fiction seemed to spill onto the pavement of a public park in China’s capital, offering a tangible glimpse into a future of human-machine interaction.
The setting itself was significant. The race unfolded in Beijing’s Yizhuang district, formally known as the Beijing Economic-Technological Development Area (BDA).40 For those unfamiliar, think of Yizhuang as a burgeoning counterpart to Silicon Valley, but with a distinct Chinese characteristic: a heavy emphasis on advanced manufacturing, hardware, and robotics, nurtured by substantial government support and strategic planning.40 Nanhai Zi Park, a large public green space, ensured maximum visibility, transforming a technological trial into a public spectacle.1 Staging such a complex and potentially risky event 27 in full public view, alongside thousands of citizens, wasn’t merely about testing circuits and code. It represented a deliberate decision by the organizers – prominent Beijing government bodies 4 – to showcase China’s technological strides directly to its people and the world, aiming to generate excitement, attract investment, and signal national capability in a cutting-edge field.5
The race promised drama: the whirring of servos replacing the pounding of feet, the potential for mechanical triumph and failure, the visible strain on batteries and balance systems. More than just a novelty, this “human-robot co-run” offered a unique window into the current state of humanoid robotics and revealed much about China’s formidable ambitions in artificial intelligence and automation.
Beneath the surface-level spectacle, the 2025 Beijing Yizhuang Half Marathon暨人形机器人半程马拉松 served multiple strategic purposes, encapsulated by its theme: “亦马当先,智领未来” (Yima Dangxian, Zhi Ling Weilai), roughly translating to “Yizhuang Marathon Takes the Lead, Intelligence Guides the Future”.4 Organized jointly by the Beijing Municipal Bureau of Sports, the Beijing Municipal Bureau of Economy and Information Technology, and the Beijing Economic-Technological Development Area (Yizhuang) Management Committee 4, the event was explicitly designed as far more than just a race.
Organizers framed it as a “top-level stress test” for humanoid robot technology.23 Pushing these machines to navigate a real-world, 21-kilometer course tested their endurance, motion control algorithms, and ability to adapt to varied environments in ways controlled lab settings cannot.5 Beyond technical validation, the goals were manifold: to showcase the potential applications of these robots 23, stimulate capital investment in the burgeoning field 23, foster collaboration among companies, universities, and research institutions 1, accelerate the pace of technological iteration 1, and crucially, to engage the public and increase understanding and interest in robotics and AI.5
The choice of Yizhuang as the venue was deliberate. This national-level economic development zone 40 has been cultivated as a nerve center for China’s high-tech ambitions, particularly in robotics and AI.40 It hosts major tech players like Xiaomi and JD.com (known for its logistics robots) and, critically, is home to the Beijing Humanoid Robot Innovation Center (BHRI) – the developer of the race’s eventual winner.42 Yizhuang aspires to be a “global highland” for the industry, with local authorities actively creating supportive ecosystems and application scenarios.4
This local focus fits within a much larger national strategy. The marathon served as a very public progress report on China’s ambitious goals for the humanoid robot sector, outlined by the Ministry of Industry and Information Technology (MIIT). These national guidelines aim to establish a preliminary humanoid robot innovation system and achieve breakthroughs enabling mass production by 2025, with the further goal of establishing a secure industrial ecosystem and achieving world-leading capabilities by 2027.50 This aligns with the broader “14th Five-Year Plan for the Robot Industry” (2021-2025), which targets making China a global center for robot innovation, manufacturing, and application, seeking annual industry revenue growth exceeding 20% and a doubling of robot density in manufacturing.56 The timing of the marathon in April 2025, just ahead of the MIIT’s target date, underscores its role as a high-profile milestone – both a catalyst pushing development forward (“以赛促研” – using competition to promote research 5) and a public demonstration of intent.
The event vividly illustrated a hallmark of China’s approach to developing strategic technologies: a powerful combination of top-down government direction, significant state investment (evidenced by MIIT plans and Yizhuang’s focused support 5), and the active participation of corporations and academic institutions (seen in the BHRI consortium structure 25 and the involvement of university teams 1).
The starting line presented a fascinating tableau of robotic diversity. Twenty teams, hailing from innovation hubs across China including Beijing, Shanghai, Guangdong, Jiangsu, and even Ningxia, brought their creations to the challenge.1 The robots varied dramatically in size, weight, and design philosophy.5 Spectators saw towering figures resembling adults standing 1.8 meters tall alongside compact machines barely reaching 75 centimeters, evoking images of metallic toddlers.5 This variety wasn’t just cosmetic; it reflected different engineering approaches to the fundamental challenges of bipedal locomotion, stability, and endurance, indicating that China’s humanoid robotics field is very much in an active, exploratory phase, with different groups pursuing distinct technological paths.
Among the notable competitors were:
Other mentioned participants included the CASBOT SE with powerful onboard computing 22, the lightweight “Elf 2” (精灵2) 22, the experimental XA1-Lite 22, the stable “Little Sea” (小海) 22, and the female-appearing “Phantom” (幻幻), which unfortunately stopped early in the race.8
To provide a clearer overview, here’s a summary of some key competitors:
Selected Humanoid Robot Competitors at the 2025 Beijing Yizhuang Half Marathon
Robot Name | Developer/Team | Height | Weight | Key Technology/Feature | Race Result/Note |
Tiangong Ultra | Beijing Humanoid Robot Innovation Center (BHRI) | 1.8m | 52-55kg | High-power joints, fast (12km/h peak), lightweight design, heat management, backed by Xiaomi/UBTECH tech 1 | 1st Place (2h 40m 42s).1 Fell once, recovered; 3 battery swaps reported.20 |
Songyan N2 | Songyan Power (Xiaowantong Team) | 1.2m | 30kg | Agile, stable, high theoretical speed (3.5m/s), affordable, ran independently? Used shoes/cooling 14 | 2nd Place (3h 37m 50s, incl. 10min penalty).17 |
Xingzhe Erhao | Shanghai Zhuoyide Robot Co. | 1.7m | 28kg | Ultra-lightweight, tendon-driven bionic design for high efficiency/endurance 38 | 3rd Place (4h 25m 56s, incl. 20min penalty).17 Reports conflict on battery swaps vs robot swaps. |
Kuafu | Tsinghua University | ~1.7m? | (Not specified) | Adapted from non-running robot, student-developed algorithms for stable gait 1 | Participated, showcased academic innovation.1 |
Xiaoju Ren | Beijing Vocational Univ. of Sci. & Tech. | 0.75m | 10kg | Smallest robot, developed by teachers/students, “social butterfly,” ambitious future plans 5 | Participated, represented grassroots involvement.5 |
Unitree G1 | Various independent teams | ~1.28m? | ~35kg? | Commercially available platform, known agility [50 (G1 specs inferred)] | Used by multiple teams; reports of falls/recovery; Unitree denied official entry.8 |
Note: Some specifications are approximate or based on related models.
The 21.0975-kilometer course itself was designed to be a formidable test.12 Starting at the South Gate of Nanhai Zi Park and finishing at the Tongming Lake Information City 1, the route traversed a mix of terrains deliberately chosen to push the robots’ capabilities. It included not only smooth asphalt but also sections with cracks and uneven surfaces, long gentle inclines, short steep slopes (up to 9 degrees reported 21), stone pathways, grass patches, and gravel areas.1 Added to this were 14 turns (six left, eight right, with minimum 90-degree angles) requiring precise navigation and balance adjustments, culminating in a final 1.5-kilometer straightaway designed to test sprint endurance.1 For bipedal robots, maintaining stability, managing power consumption, and perceiving and adapting to these varied conditions over such a distance presents immense engineering challenges.12
Recognizing these difficulties, the organizers established specific rules tailored for the robotic participants, balancing the desire for a challenging test with the realities of current technology:
These rules, particularly the allowances for battery swaps, relays, and the extended cut-off time, clearly reflect a pragmatic understanding by the organizers of the current technological state. They prioritized enabling participation and gathering valuable real-world performance data over enforcing rigid adherence to human marathon standards. This pragmatic approach underscores the event’s primary function as a demanding, public testbed designed to push the boundaries of the technology.7
The inherent challenges were numerous. Beyond navigating the course, robots had to maintain stability while dealing with the constant impact forces generated by running.1 Endurance was a major factor, not just in terms of battery life but also mechanical reliability over tens of thousands of steps.23 Heat management, especially in the joints, proved critical, as prolonged exertion can cause components to overheat and fail.1 Teams employed various strategies to cope, including lightweight designs 1, specialized heat dissipation techniques 1, external cooling sprays 14, and even having their robots wear athletic shoes to cushion impact and improve grip.12
As the starting gun fired at 7:30 AM, the robots began their historic journey, departing sequentially at one-to-two-minute intervals to ensure safe spacing.21 The 1.8-meter Tiangong Ultra was the first off the line, its long metallic legs propelling it forward at a determined pace, quickly establishing a lead.1 Close behind followed the smaller but nimble Songyan N2 robot from the Xiaowantong team, its shorter legs churning rapidly.20 The crowd responded with cheers and gasps, witnessing the start of this unprecedented race.5
The ensuing hours provided a compelling narrative of both technological prowess and unexpected fragility. Tiangong Ultra maintained a remarkably consistent pace, averaging between 7 and 8 kilometers per hour for much of the race.5 As anticipated, its high performance demanded significant energy; the team executed its first battery swap around the 5km mark, taking approximately 1 minute and 50 seconds.20 In total, Tiangong reportedly required three battery swaps during the race.22 The most dramatic moment came roughly two hours in, around the 17km mark, when the leading robot suddenly lost balance on a narrow straight section and tumbled to the ground.20 Engineers rushed to its side. Post-race analysis suggested the fall was caused by an inaccurate battery level reading, leading to unexpected power loss.20 Impressively, after rapid adjustments by the team, Tiangong Ultra managed to push itself back up and resume running, demonstrating a resilience that resonated with the human marathon spirit.5 This sequence – “Fall, get up, swap battery, set off again” 5 – became a defining moment, showcasing both the robot’s capability and its current limitations.
Meanwhile, the Songyan N2 robot ran a steady race, noted for its stability and, according to some reports, a higher degree of autonomy, navigating significant portions without close human accompaniment.14 Its team employed practical solutions like children’s running shoes and cooling spray to manage the rigors of the course.14
The Xingzhe Erhao robot presented a different kind of endurance story. While its overall pace was slower, its highly efficient tendon-driven design potentially allowed it to cover the distance with fewer energy demands. Initial reports celebrated it as the only robot to finish without needing a battery swap.73 However, the official results showing two robot-swap penalties 17 create ambiguity. It’s possible the team opted to swap the entire lightweight robot rather than just the battery, or there may be conflicting accounts of its race strategy. Regardless, its performance highlighted the trade-offs between speed and energy efficiency.
The course proved too much for many contenders. Falls were common 1, and some robots suffered mechanical failures, including one widely circulated anecdote of a robot losing its “head” but continuing to run briefly before being attended to.14 Others, like the female-form robot Huanhuan, stopped early on.8 Ultimately, only six of the 20 starting teams officially completed the full 21.0975 km distance within the cut-off time.5 This low completion rate starkly illustrated the immense difficulty of the challenge. Organizers and participants adopted a philosophical stance, emphasizing that every attempt, even unsuccessful ones, represented “effective failure” (有效失败) 18 – valuable learning experiences. The mantra “Participation is victory, finishing makes you a hero” (参赛即胜利,完赛即英雄) echoed this sentiment.9
Throughout the race, the human element was crucial. Engineers were constantly monitoring, adjusting, and intervening – rushing to perform rapid battery swaps under pressure 7 or quickly diagnosing and fixing problems after a fall.7 Human guides carefully paced their robotic partners 1, while the crowds lining the route offered encouragement not just to the human runners, but to every determined stride of the metallic competitors.7
Just after 10:00 AM (reports vary slightly between 10:10 and 10:14 AM 1), cheers erupted as Tiangong Ultra crossed the finish line.1 Its official time was recorded as 2 hours, 40 minutes, and 42 seconds 1, a time hailed as a world record for a humanoid robot completing a half marathon under such conditions.8 Following Tiangong were the two Songyan N2 teams: Xiaowantong secured second place with a final time of 3 hours, 37 minutes, and 50 seconds (base time 3:27:50 plus a 10-minute penalty for one robot swap).17 The Xuanfeng Xiaozi team, also using an N2, reportedly took third, though their exact time isn’t consistently reported across sources but likely involved penalties as well. Xingzhe Erhao finished later, its official time being 4 hours, 25 minutes, and 56 seconds, reflecting a base time of 4:00:56 plus 20 minutes in penalties for two robot swaps.17 The finishers were awarded unique robot-shaped medals, commemorating their participation in this historic event.17
The significant time gaps between finishers, coupled with the high attrition rate, underscored the reality that sustained, dynamic bipedal locomotion in complex, real-world environments remains a frontier challenge for robotics. While Tiangong’s speed was impressive, the need for multiple battery swaps contrasted with Xingzhe Erhao’s potential (though debated) endurance, highlighting that “performance” can be measured in different ways – speed versus efficiency – and the official ranking might not capture the full spectrum of technological progress demonstrated.
The Beijing Humanoid Robot Half Marathon was far more than an isolated technological feat; it was a carefully orchestrated event deeply embedded within China’s national strategy for technological advancement. It served as a high-profile, public demonstration – likened by one official to a “科技阅兵式” (technology parade) 20 – showcasing progress towards the ambitious goals set forth in the MIIT’s guidelines for humanoid robot development (mass production by 2025, global leadership by 2027) 50 and the broader objectives of the 14th Five-Year Plan for the Robot Industry.56
The choice of Beijing’s Yizhuang district was instrumental. This area has been meticulously developed as a specialized hub designed to cultivate exactly this type of innovation, boasting a complete robotics industry chain from core components to system integration and application scenarios.40 Central to this ecosystem is the Beijing Humanoid Robot Innovation Center (BHRI), the entity behind the winning Tiangong Ultra robot. BHRI itself exemplifies China’s collaborative approach; it’s a consortium uniting major private tech companies like Xiaomi and UBTECH with state-linked enterprises (Jingcheng Electromechanical) and the local government’s Yizhuang robotics platform.25 This structure is designed to pool resources, share expertise, and tackle fundamental technological hurdles collaboratively, potentially accelerating progress beyond what individual companies might achieve alone. Tiangong’s victory, explicitly leveraging technology contributions from Xiaomi (AI vision/control) and UBTECH (joint tech/reliability) 49, serves as an early validation of this public-private partnership model. BHRI’s recent upgrade to a national-level innovation center 23 and associated plans for large-scale funding initiatives (like a potential 10 billion RMB fund 61) further signal the significant strategic and financial backing behind this approach. The local Yizhuang government complements this by actively creating real-world application opportunities, such as its “10,000 Robot Plan” which includes specific demand for nearly a thousand humanoid robots across various sectors.38
The marathon itself acts as a powerful catalyst for the industry. The intense pressure of competition drives innovation (“以赛促研” – use competition to promote research 5) and forces rapid iteration.1 It provides a platform for teams to observe each other’s designs and strategies, fostering informal knowledge sharing and collaboration.1 Furthermore, the high visibility of the event helps attract investment, talent, and public interest, creating a more vibrant ecosystem.2 Beijing’s plans to host a “World Humanoid Robot Games” featuring a wider range of robotic athletic competitions later in 2025 23 indicate a sustained commitment to using competitive events as a development accelerator.
This strategy plays to China’s inherent strengths in this field: a comprehensive and robust manufacturing supply chain capable of producing the thousands of components needed for complex robots 5, a vast domestic market offering diverse potential application scenarios 38, strong and directive government support through national and local planning 5, and a rapidly growing pool of engineering and AI talent.5 The marathon, therefore, wasn’t just a race; it was a microcosm of China’s determined, multi-faceted strategy to achieve leadership in a critical future technology.
While the sight of robots running a marathon was undeniably impressive and hailed as a “historic moment” 1, expert commentary and the race’s unfolding drama provided a sober assessment of the technology’s current maturity. The event served as a crucial public benchmark, revealing both significant advancements and persistent limitations.33
Organizers and industry insiders consistently emphasized the marathon’s role as a test, urging patience and a tolerance for the iterative nature of technological development (“包容的心态看待新技术迭代” – view new tech iteration with tolerance 33; “一步一个脚印” – one step at a time 5). The numerous technical difficulties encountered during the 21km ordeal underscored the major hurdles that remain before humanoid robots can be reliably deployed for widespread practical use:
Despite these challenges, the vision for humanoid robots is expanding rapidly. Industry leaders speak of moving robots from “stage performances” to practical work in factories (“工厂里用起来”).33 Potential applications frequently cited include roles in advanced manufacturing, logistics and warehousing, healthcare and elder care, home services, agriculture, and performing tasks in hazardous or inaccessible environments.5 The developers of Tiangong are already planning “Tiangong 2.0,” which will feature enhanced dexterity with the addition of functional hands, aiming for better upper and lower body coordination and manipulation capabilities.49
Globally, the Beijing event garnered significant attention, albeit with mixed reactions. International media coverage and online commentary reflected a blend of awe at the technological display, amusement at the occasional mishaps, and a degree of underlying unease about the implications of increasingly capable humanoid robots.30 Comments ranged from comparisons to science fiction films like “I, Robot” 81 to descriptions like “a bit terrifying” 82 and questions about potential job displacement.81 This global fascination underscores the fact that humanoid robotics, particularly with China’s visible advancements, is crossing a threshold into mainstream consciousness, raising both excitement and complex societal questions. China, meanwhile, has clearly stated its ambition not just to participate, but to achieve a globally leading position in this transformative technology 50, setting the stage for continued international focus and competition.
Lost amidst the focus on circuits and sensors is the undeniable human element that powered this robotic endeavor. The marathon wasn’t just a test of machines; it was a testament to the ingenuity, perseverance, and collaborative spirit of the engineers, students, and researchers behind them.
University teams played a prominent role, showcasing the pipeline of talent and innovation emerging from China’s academic institutions. The undergraduate team from prestigious Tsinghua University faced the daunting task of modifying their “Kuafu” robot, not originally built for running, developing novel algorithms on the fly to achieve stable locomotion.1 Equally inspiring was the “Xiaoju Ren” (Little Giant) team from Beijing Vocational University of Science and Technology. Comprising seven faculty members and four students, they built their diminutive robot from the ground up as an extracurricular project, driven by a long-term vision of incremental improvement year after year.5 These examples highlight the crucial role universities play not just in fundamental research but also in hands-on application and cultivating the next generation of roboticists.46
The pressures on the corporate engineering teams were immense on race day. Videos and reports captured moments of intense focus during rapid-fire battery swaps 7 and the quick thinking required to diagnose and fix robots that had stumbled or malfunctioned mid-race.7 The human guides, like Jia Ning who partnered with Tiangong Ultra, developed a unique working relationship with their robotic counterparts, learning to anticipate their movements and guide them effectively through the course.65 His experience, transitioning from a marathon enthusiast to a robot’s lead runner, involved weeks of practice to synchronize pace and movement, highlighting the novel forms of human-machine collaboration emerging.65
The visions of the company founders also shone through. Professor Li Qingdu of Zhuoyide explained the thoughtful meaning behind naming their robot “Xingzhe” (Walker) – emphasizing a steady, step-by-step approach to overcoming the significant challenges still facing the field.38 He also stressed the importance of industry collaboration for mutual learning and advancement.1 Songyan Power’s founder, Jiang Zheyuan, expressed surprise and pride at his young company’s N2 robots securing second and third place, humorously pitching their affordability by claiming a four-day return on investment.14
Beyond the technical achievements, the race was filled with memorable, almost anthropomorphic moments: robots waving to the crowd 5, the slightly comical sight of robots sporting running shoes 14, the startling incident of a robot losing its head but attempting to soldier on 14, and the sheer determination displayed by robots programmed to get back up after a fall and continue towards the finish line.5 These moments, coupled with motivating slogans from organizers like “The marathon has an end point, but the innovation and development of humanoid robots do not” 1 and “Humanoid robot ‘running a marathon’ is essentially human dreams running” 5, captured the spirit of the event.
The collaborative undercurrent was also palpable. Teams were observed studying each other’s designs and performance 1, embodying the idea that shared competition and learning benefit the entire industry’s growth.1 The mention of specialized companies like Virtual Point providing crucial gait data services for participants further points to an emerging ecosystem of support.49 The public framing of the event, celebrating participation (“Participation is victory” 9) and valuing the lessons learned from failures 18, fostered an atmosphere conducive to continued experimentation and collective progress, aligning with the broader goal of building a robust national industry.
The 2025 Beijing Yizhuang Humanoid Robot Half Marathon has undoubtedly etched its place in the annals of technology. As the first event of its kind, pitting advanced humanoid robots against the endurance challenge of a half marathon alongside thousands of human runners, it offered a compelling, and at times dramatic, snapshot of the state-of-the-art in Chinese robotics.1 It was a bold statement of ambition and a revealing public test.
The results showcased impressive advancements, particularly in dynamic bipedal locomotion, balance recovery, and the integration of complex hardware and software systems. Robots like Tiangong Ultra demonstrated the ability to run at sustained speeds for extended periods, while others like the Songyan N2 showed remarkable agility, and the Xingzhe Erhao highlighted potential breakthroughs in energy efficiency. The event successfully served its purpose as an extreme “stress test,” pushing boundaries and generating invaluable data for future development.7
However, the marathon also laid bare the significant hurdles that remain. Issues with battery endurance, mechanical reliability under stress, joint overheating, and the persistent need for human oversight for navigation and control highlighted the gap between current capabilities and the requirements for widespread, autonomous deployment in complex real-world scenarios.33 The low completion rate served as a stark reminder that the path from laboratory potential to robust practical application is long and challenging.
Yet, the commitment is clear. Organizers are already planning future iterations of the event, aiming for broader participation, faster robots, improved environmental adaptability, and greater autonomy.5 This, combined with national strategic plans, dedicated innovation hubs like BHRI, and the upcoming World Humanoid Robot Games 23, signals China’s unwavering determination to lead in this field.
The marathon also offered a fascinating glimpse into the nascent stages of “人机共生” (human-machine coexistence) and “人机共荣” (human-machine co-prosperity).5 Seeing humans and humanoids sharing the same starting line, navigating the same city spaces (albeit on separate tracks), and striving towards a common finish line, however imperfectly, sparked imagination and debate about the future integration of these intelligent machines into society.
The finish line at Tongming Lake has been crossed, but as Beijing Economic-Technological Development Area official Li Quan aptly stated, “The marathon’s finish line is crossed, but the exploration of human-machine collaboration will not stop”.5 For humanoid robotics, the real marathon – the long run towards widespread utility and seamless integration into human life – has only just begun. And China has decisively positioned itself as a formidable contender in that race.
After a major restructuring, TikTok's parent company, ByteDance, is doubling down on gaming. Explore their…
Burned out by the relentless pressure to succeed? Discover "Tang Ping," the Chinese youth movement…
Discover the incredible story of IFBH, the Thai company behind "if" coconut water, which achieved…
Explore how soaring cemetery prices in China are challenging ancient burial customs, leading to phenomena…
Uncover the dramatic story of Zhang Xiaoquan, China's iconic 400-year-old knife maker, now battling financial…
Discover Xiao Zhi AI, the open-source Chinese project taking the AI hardware world by storm.…