Xynova Flex 1

RawGraph

Last edited

Fact-checked

In review queue

Sources

18 citations

Revision

v4 · 2,570 words

Fact-checks are independent of edits: a reviewer re-verifies the article against its sources and stamps the date. How we verify

Xynova Flex 1
MakerXynova (Hangzhou, China)
TypeHigh-DOF tendon-driven dexterous hand
Announced2025 (June, per the company); shown at CES 2026
Active DOF20
Total DOF25 (20 active + 5 passive)
Palm weightAbout 380 g
Whole-hand graspMore than 30 kg
DrivePure tendon-driven

The Xynova Flex 1 is a high-degree-of-freedom tendon-driven dexterous hand built by Xynova (Chinese: 曦诺未来), a Hangzhou robotics startup founded in December 2024. Released in 2025, it packs 25 total degrees of freedom (20 active plus 5 passive) into a palm assembly weighing about 380 grams, while the company rates its whole-hand grasp load at more than 30 kilograms. [1][3] Xynova bills the Flex 1 as the world's first mass-produced high-DOF hand driven purely by tendons, and a 2026 dexterous-hand supply-chain infographic named the "Hangzhou Xynova Flex 1" as the representative tendon-driven product in its category. [1][6]

What is the Flex 1?

The Flex 1 is Xynova's first-generation product and the hand that established the company as a serious component-layer supplier for humanoid robots and embodied AI systems. Rather than sell a full robot, Xynova sells the end-effector and the actuator stack inside it, and the Flex 1 is the showcase for that stack. It is an arm-hand integrated unit: the five-fingered hand is paired with a forearm module, and most of the actuation lives in the forearm rather than in the hand itself. This is the same anatomical trick the human body uses, where large muscle groups in the forearm pull on tendons that run down into a light, low-inertia hand. [4]

Because the motors and screws sit in the forearm and drive the fingers through cables, the palm assembly stays light (about 380 g) while the whole hand can still generate high force. Founder Xia Yuxuan has described the design as a first-principles attempt to beat the dexterous hand's "impossible triangle," the industry shorthand for the fact that degrees of freedom, compact size, and force output pull against one another and cannot all be maximized at once. The Flex 1 deliberately pushes the DOF-and-force corner of that triangle. [4][3]

The company positions the hand for research groups, teleoperation rigs, and integrators who want the largest possible mechanical envelope (many independently moving joints plus heavy grasp load) in a hand-sized package. [3][4]

Specifications

Reported figures are drawn from Xynova's product materials and Chinese-language technical coverage. Vendor performance claims are attributed and should be read as manufacturer specifications rather than independently benchmarked results.

ParameterXynova Flex 1
Active degrees of freedom20
Total degrees of freedom25 (20 active + 5 passive)
Wrist degrees of freedom2 (part of the arm-hand module)
Palm weightAbout 380 g
Whole-hand grasp / lift loadMore than 30 kg
Fingertip forceMore than 20 N per finger
Full open-close cycleAbout 0.6 s
Repeatability0.2 mm or better
TransmissionPure tendon-driven (steel cables)
ActuatorsIn-house micro electric cylinders, roughly 10 to 12 mm, 100 to 300 N thrust
Motors8 mm hollow-cup brushless motors
Screws7 mm planetary roller screws
Tendon durability (company claim)More than 1,000,000 open-close cycles at rated load

The load figures are the Flex 1's headline numbers: at about 380 g of palm weight and a stated grasp load above 30 kg, Xynova argues the hand has an unusually high force-to-weight ratio, which matters for battery-limited mobile humanoids. [1][3][10] The combination of a fast full-hand cycle (about 0.6 s) with sub-0.2 mm repeatability and 20-plus newtons of fingertip force is what the WRC-era technical coverage highlighted as its standout profile. [3]

One caveat on the numbers: early promotional material from June 2025 described the hand as having 24 DOF and weighing about 300 g, but the settled specification carried across Xynova's own product page and later coverage is 25 total DOF (20 active plus 5 passive) and about 380 g of palm weight. [9][1][2] Some reseller listings copy generic template values (one lists "6 active DOF") that conflict with the manufacturer's figures and should not be relied on. [11]

The pure tendon-driven design

The Flex 1's defining choice is a pure tendon-driven transmission. Every active joint is moved by a steel cable that behaves like an artificial muscle-tendon, pulled by an actuator seated in the forearm rather than by a motor built into the finger. Xynova contrasts this with the linkage-driven and gear-driven hands common among competitors, and argues that routing all actuation through cables is what lets it fit 20 active joints into a hand this light. [1][3]

What makes the approach practical is vertical integration of the actuator stack. Xynova designs and manufactures the whole electromechanical chain in-house: roughly 8 mm hollow-cup brushless motors, 7 mm planetary roller screws, the 10 to 12 mm micro electric cylinders that combine them (rated at 100 to 300 N of thrust), the reducers, and the control algorithms. [1][3] These micro-actuators and roller screws are typically the supply bottleneck for Chinese hand makers, so producing them internally is the company's core differentiator and the reason it can talk about mass production at all. The company summarizes the stack as "motors, electronic controls, reducers, screws, and algorithms." [3][4]

The trade-off is precision. A pure-tendon layout maximizes the number of independently actuated joints and the peak force per unit of hand weight, but cables stretch under load and introduce friction and hysteresis, which cap how finely the hand can hold a position or modulate force. This is precisely why Xynova's second-generation hand, the Flex 2 (launched May 2026), moved to a hybrid drive that mixes cable-driven joints with direct-drive joints: the direct-drive elements tighten repeatability to better than plus or minus 0.1 mm and force control to about 0.05 N, at the cost of some raw grasp load. In other words, the Flex 1 optimizes DOF and force output, and the Flex 2 trades some of that away for finesse and sensing. [7][8]

Xynova also stresses durability, a chronic weak point of tendon hands, whose cables can fray or creep. The company says the Flex 1's tendon transmission was validated past 1,000,000 open-close cycles at rated load, and it markets anti-creep tendon ropes plus full temperature, shock, and EMC validation for industrial deployment. [1][4] In a 2026 interview it reported component-level testing of about 1.8 million cycles on tendon parts and more than 5 million cycles on the micro electric cylinders. [4] These vendor figures sit far above the tendon durability framing used in the 2026 industry infographic, which described a target failure rate under 0.1 percent over 10,000 open-close cycles; the infographic's threshold and Xynova's own million-cycle claim are different kinds of statement and should not be conflated. [6][4]

Sensing and control

The Flex 1 is primarily an actuation and mechanics showcase, and its sensing is more modest than what Xynova later built into the Flex 2. The Flex 1 supports both position and force control, and its back-drivable actuators allow a degree of force feedback and compliant grasping, which is what lets it adjust grip on deformable objects like paper and bags in demonstrations. [3][4] Its two wrist degrees of freedom let the hand reposition and reorient (Xynova has demonstrated it writing with a pen) without moving the whole arm. [3]

What the Flex 1 does not advertise is the richer multimodal perception stack that arrived with the Flex 2. The multimodal "cerebellum"-style control layer, the wrist-mounted camera, and the fused force, tactile, visual, and proximity sensing used for slip detection and adaptive grasping are Flex 2 features, and Xynova has not published equivalent full-hand tactile-skin specifications for the Flex 1. [7][8][1] Readers comparing the two generations should treat the advanced sensing narrative as belonging to the Flex 2.

Manufacturing and availability

Xynova is scaling the Flex 1 into volume production rather than treating it as a demonstrator. The company is building a facility of roughly 5,400 square meters designed for an annual run rate of about 10,000 high-DOF dexterous hands and 200,000 micro electric cylinders, with ramp-up through the second quarter of 2026 and full capacity targeted by year-end. [4][6] Before the line even entered commissioning, Xynova said it had secured orders for more than 10,000 hands from a leading humanoid-robot manufacturer. [4][6] The Flex 1 also appears on open commercial marketplaces, listed for sale as a 25-DOF, 380 g, greater-than-30 kg-load humanoid hand, which underlines that it is a purchasable product rather than a lab prototype. [10]

The hand's public profile grew through 2025 and into 2026. It was covered internationally by the RoboHub account in September 2025 as Xynova's first fully self-developed high-DOF tendon hand, and it was demonstrated at CES 2026 in January, where coverage framed it as one of the mass-produced Chinese dexterous hands drawing attention on the show floor. [2][7][8] Xynova's rapid fundraising (nearly 1 billion yuan across three rounds by May 2026, from backers including Xiaomi, CATL, and JD.com) is detailed on the Xynova company page. [5][12]

Positioning: Flex 1 versus Flex 2 and peer hands

The Flex 1 sits at the high-force, high-DOF extreme of the robot manipulation hardware market. Against its own successor and against peer Chinese and captive hands, its distinguishing traits are the lightest palm weight and by far the highest stated grasp load in the high-DOF class, achieved through the pure-tendon architecture. The following table compares published manufacturer figures; drive types and load definitions vary between vendors, so the numbers are directional rather than strictly like-for-like. [3][4][14][15][16][17][18]

HandDOFDriveHand / palm weightGrasp load (company figure)
Xynova Flex 125 total (20 active + 5 passive)Pure tendon-drivenAbout 380 gMore than 30 kg
Xynova Flex 223 total (19 active + 4 passive)Hybrid (cable + direct drive)About 400 g12 kg single-hand peak
Inspire Robots RH566 active (12 joints)LinkageAbout 500 to 790 g by variantRoughly 3 to 5 kg
Linkerbot LinkerHand L30About 21 to 22 (17 active)TendonAbout 1.4 kg5 kg payload (12 N grip)
Wuji Hand20 activeDirect-driveAbout 580 g10 kg static, 5 kg vertical
ZWHAND17 to 20 activeIn-finger linear motors (direct)Not publishedNot published
Unitree Dex520 total (16 active + 4 passive)Tendon plus linkageAbout 1 kgA few kilograms

Two patterns stand out. First, on the weight-versus-grasp trade, the Flex 1 is an outlier: it is lighter than the Wuji Hand, the Unitree Dex5, and the Linkerbot L30 while claiming several times their grasp load, which is the direct payoff of moving the actuators off the hand and into the forearm. [3][14][15][16] Second, the higher-DOF peers cluster around finer precision or richer sensing (the Wuji Hand's direct-drive rotary joints, the Unitree Dex5's optional 94-point tactile array, ZWHAND's in-finger linear motors and e-skin) rather than raw force, which is the corner of the design space the Flex 1 gives up. [14][16][17]

Reception

Coverage of the Flex 1 has been broadly positive on its mechanical achievement while appropriately cautious about unverified marketing figures. Chinese technical outlets treated it as a genuine engineering result rather than a slideware demo, emphasizing that the hand had already gone through multiple hardware iterations and entered production, and singling out the light palm, high grasp load, and in-house actuator stack as the differentiators. [3][4] International robotics accounts framed the Flex 1 as evidence that the hard, unglamorous component layer (the hands, motors, and screws that decide what a robot can actually do with objects) was becoming a competitive battleground, and its CES 2026 showing was described as one of the standout mass-produced Chinese hands. [2][7][8] As with all vendor-supplied figures, the durability and load claims here await independent third-party benchmarking.

ELI5

Imagine a robot hand that is as light as a small apple but can hold up something as heavy as a big bag of dog food. The Xynova Flex 1 does this with a clever trick borrowed from your own body. Your fingers do not have muscles inside them; the muscles are up in your forearm, and they pull on thin cords (tendons) that make your fingers move. The Flex 1 works the same way: the little motors are hidden in the wrist-and-forearm part, and thin steel cables run down to the fingers and yank them, so the hand itself stays light. That lets it have 20 moving joints and still lift a lot. The downside is that pulling on long cables makes it a little less precise, which is why Xynova's next hand, the Flex 2, added some motors right at the joints to make it steadier and give it a better sense of touch.

See also

References

  1. "Xynova Flex 1." Xynova official product page, 2026. https://www.xynova.com.cn/en/xynova-flex-1
  2. RoboHub, X post on the Xynova Flex 1 (380 g, 25 DOF, 20 active plus 5 passive), September 9, 2025. https://x.com/XRoboHub/status/1965332822710972524
  3. "25 DOF and 20 N Fingertip Force: WRC's Strongest Humanoid Dexterous Hand Emerges" (25自由度+20N指尖力,WRC最强仿人灵巧手横空出世). Leaderobot / Robot Lecture Hall (机器人大讲堂), 2025. https://www.leaderobot.com/news/6049
  4. "Interview with Xynova: Cracking the Dexterous Hand's 'Impossible Triangle' from First Principles" (专访曦诺未来Xynova|从第一性原理出发,破解灵巧手 "不可能三角"). DoNews, April 2026. https://www.donews.com/article/detail/8280/99220.html
  5. "Chinese Robotics Startup Xynova Raises Over US$13 Million in Angel Round for High-DOF Dexterous Hands." TMTPost, December 26, 2025. https://en.tmtpost.com/post/7822011
  6. "From Prototypes to Production: Dexterous Hands Kick Off a Mass-Production Race." Gasgoo Auto News, 2026. https://autonews.gasgoo.com/articles/news/from-prototypes-to-production-dexterous-hands-kick-off-a-mass-production-race-2016425582734970881
  7. "First Look: Xynova Flex 1, The Mass-Produced Dexterous Hand Steals the Show at CES 2026." YouTube, January 2026. https://www.youtube.com/watch?v=xHg8lodr_II
  8. "CES 2026: Eastern Robots Punch the West Coast" (CES 2026,东方机器人拳打西海岸). OFweek AI, January 2026. https://www.ofweek.com/ai/2026-01/ART-201700-8110-30678804.html
  9. Discover Yuhang, X post on the Xynova Flex 1 (early 24-DOF / 300 g framing), June 2025. https://x.com/DiscoverYuhang/status/1935245338249674786
  10. "Xynova Flex 1 Lightweight 380g Load 30kg 25 DOF Humanoid Robotic Hand." Alibaba product listing, 2026. https://www.alibaba.com/product-detail/Xynova-Flex-1-Lightweight-380g-Load_1601687364959.html
  11. "Xynova Flex 1 Dexterous Hand with 25 DoF for Precise Manipulation and Human-Like Motion." Gongboshi Robot marketplace listing, 2026. https://www.gongboshi-robot.com/sale-54519390-xynova-flex-1-lightweight-380g-load-30kg-maximum-thrust-250n-25-dof-humanoid-robotic-hand-dexterous-.html
  12. "Dexterous-Hand Company Xynova Future Completes Angel Round of Over 100 Million Yuan" (灵巧手企业「曦诺未来 Xynova」完成超亿元天使轮融资). 36Kr, December 2025. https://36kr.com/p/3611715966104072
  13. "About Us." Xynova official site, 2026. https://www.xynova.com.cn/en/about-us
  14. "All You Need to Know About the Unitree Dex5 Hand." Humanoid.guide, 2025. https://humanoid.guide/all-you-need-to-know-about-the-unitree-dex5-hand/
  15. "Linkerbot Linker Hand L30." AIFITLAB product page, 2025. https://aifitlab.com/products/linkerbot-linker-hand-l30
  16. "Wuji Hand: Product Introduction." Wuji Technology documentation, 2025. https://docs.wuji.tech/docs/en/wuji-hand/latest/overview/
  17. "ZWHAND Dexterous Hand." ZWHAND (Shenzhen Zhaowei) official site, 2025. https://www.zwhand.com/en
  18. "The Dexterous Hands RH56DFX Series." Inspire Robots, 2024. https://en.inspire-robots.com/product/rh56dfx

Improve this article

Add missing citations, update stale details, or suggest a clearer explanation. Every suggestion is reviewed for sourcing before it goes live.

3 revisions by 1 contributors · full history

Suggest edit