Dexterous robot hands have become one of the more closely watched races in physical AI, and Hangzhou-based startup Xynova just added a second entry to its own lineup. Having unveiled the tendon-driven Flex 2 hand at ICRA 2026 in Vienna back in May, the company is now showing a direct-drive platform called Prima1 at the World Robot Conference (WRC) in Beijing — 22 degrees of freedom, tactile sensing, and high-precision force control, aimed less at flashy humanoid demos and more at industrial manipulation work.
TL;DR
| Question | Answer |
|---|---|
| What is Prima1? | Xynova's new direct-drive dexterous hand — 22 DoF, tactile sensing, precision force control |
| How is it different from Flex 2? | Flex 2 is hybrid tendon+direct drive (23 DoF); Prima1 is pure direct drive (22 DoF) |
| What's it built for? | Industrial and research-grade manipulation, not humanoid showpieces |
| Where is it shown? | World Robot Conference, Beijing, physical hand on display |
| Who funds Xynova? | Li Auto and CSC Financial Capital led a hundreds-of-millions-yuan Series A; Meituan and Xiaomi also invested |
| How does it compare to 1X's NEO hand? | Both high-DoF, but NEO optimizes for compliant human-safe interaction; Prima1 optimizes for industrial precision |
Direct drive vs. tendon drive, and why Xynova is building both
Most dexterous hands pick one actuation approach and commit to it. Xynova is building two hands on two different architectures at the same time, and the difference explains what each one is actually for.
Tendon drive (Flex 2, and 1X's NEO hand) routes cables from motors — often positioned away from the joint itself, sometimes in the forearm — to pull fingers into position. It's lightweight at the point of contact and naturally compliant, which is why it shows up in hands meant to interact safely with people and unpredictable environments.
Direct drive (Prima1) puts a motor at or near each joint with no cable in between. That removes the backlash and elasticity tendons introduce, buying tighter, more repeatable force control — the kind of precision industrial manipulation tasks reward — at the cost of extra weight and mechanical complexity packed into the hand itself.
Xynova's own Flex 2, announced in May, actually splits the difference with a hybrid "tendon plus direct drive" design at 23 degrees of freedom. Prima1 goes further toward pure direct drive at 22 DoF, positioned by Xynova as a research-grade platform for embodied-intelligence R&D — a deliberately narrower, more precision-focused sibling to Flex 2 rather than a replacement for it.
How this stacks up against 1X's NEO hands
Xynova isn't the only company racing to ship high-DoF dexterous hands this year. 1X shipped 25-DoF tendon-driven hands for its NEO humanoid on July 9, 2026, with force-transparent joints and tactile shear sensing tuned specifically for safe operation around people in homes. That's a different design target than Prima1: 1X is optimizing for compliant, forgiving interaction as part of a general-purpose humanoid; Xynova's direct-drive line is optimizing for the kind of tight, repeatable force control that industrial assembly and manufacturing tasks need, where the hand doesn't need to be gentle with a human, it needs to be exact with a part.
That split reflects a broader pattern in the dexterous-hand market right now: nobody has settled on one actuation architecture as clearly superior, because "best hand" depends entirely on what the hand needs to do — and increasingly, companies like Xynova are choosing to build more than one hand rather than bet on a single answer.
Xynova's funding trajectory
Xynova was founded in 2024 and has moved fast on capital. An early angel round raised more than $13 million, followed by a pre-Series A round of several hundred million yuan in March 2026 with Xiaomi's venture arm increasing its stake, then a Series A of hundreds of millions of yuan led by Li Auto and CSC Financial Capital, with Meituan and Xiaomi participating again — close to RMB 1 billion raised within roughly six months, according to Chinese financial outlets covering the round. That pace of funding puts Xynova alongside the broader wave of Chinese dexterous-hand and humanoid startups that explainx.ai has tracked through China's humanoid robot training schools piece — a sign of how much capital is chasing manipulation hardware specifically, separate from the humanoid-robot bodies those hands eventually attach to.
Related reading
- 1X NEO Hands: 25-DoF Tendon Drive, Force Transparency, and 10K Units in 2026
- China's Humanoid Robot Training Schools
- YC Paper Club: Why Robotics Still Isn't Solved
- Dyna-2 World Action Model and the Robotics Scaling Law
- Matic Cues: On-Device Robot AI With Jetson Orin Nano
- Genesis AI's Eno: An Agentic Robot
Official source: Xynova (曦诺未来) · World Robot Conference 2026 exhibitor listing
Specs and funding figures reflect what was publicly reported as of August 17, 2026. Xynova has not disclosed full Prima1 performance metrics (weight, payload, fingertip force); some details may only surface at WRC itself.