# GENTO Debuts 21-DoF Dexterous Hand at CoRL 2026

Tianji Intelligence unveils advanced robotic manipulation hardware alongside NVIDIA GR00T 2 integration.

- **Source:** [Digital Trends](https://www.digitaltrends.com)
- **Published:** 2026-11-12
- **Topic:** [technology](https://brifing.si/c/technology)
- **Original article:** <https://www.digitaltrends.com/brc/gento-debuts-a-21-dof-hand-at-corl-2026-as-nvidia-launches-gr00t-2-with-tianji-marvin-pro/>
- **On Brifing:** https://brifing.si/s/3d696520-207a-4b22-b7db-2137d35ca80e

## Dexterous Hand Debut

Tianji Intelligence introduced its proprietary 21-degree-of-freedom dexterous hand at the Conference on Robot Learning 2026 in Austin. The hardware aims to solve common robotic limitations like high backlash and fragile transmission. By utilizing a linkage-driven design, the hand provides consistent movement and increased fingertip force for complex, real-world physical manipulation tasks.

## Solving Mechanical Hurdles

The GENTO hand addresses four critical engineering challenges: high backlash, fragile transmission, rigid joints, and limited fingertip force. Its design allows for backdrivability, meaning external forces can move the fingers without meeting resistance. Paired with MIT force-position hybrid control, the system enables compliant, human-safe interactions that were previously difficult to achieve reliably.

## NVIDIA GR00T 2 Integration

At CoRL 2026, NVIDIA showcased its GR00T 2 open-world action model using the TIANJI Marvin Pro platform. This demonstration featured coordinated two-handed manipulation, such as preparing banana yogurt. The platform serves as a physical execution layer, bridging the gap between digital AI policies and the messy, unpredictable nature of real-world physics.

## Advanced Sensor Technology

The foundation of the TIANJI lineup relies on in-house MEMS joint torque sensors. These sensors detect torques as low as 0.02 N m with a 160 Hz bandwidth and 0.03 mm positioning accuracy. Integrated into joint modules, they allow robotic arms to act as active sensory organs without requiring frequent, time-consuming recalibration processes.

## Data Collection Pipeline

Tianji Intelligence has developed a comprehensive workflow connecting VR headsets and data gloves to robot platforms. A fleet of 100 TIANJI Marvin Pro units can collect over 600,000 real-robot demonstration datasets annually. This data is essential for training AI models, as it captures physical resistance alongside synchronized vision and motion data.

## Mobile Platform Ecosystem

Beyond fixed arms, the company offers mobile platforms like GENTO Luna and GENTO Skye. Luna is a wheeled, foldable humanoid featuring an omnidirectional base and dual arms for real-time teleoperation. Skye serves as a dedicated development platform for reinforcement learning, providing a robust environment for validating deployments in dynamic, real-world settings.

## Industrial Market Presence

Jointly incubated by Yaskawa Electric and Everwin Precision, Tianji Intelligence has deep industrial roots. The company delivered over 2,000 force-controlled humanoid dual arms in mid-2025 alone. With over 30,000 industrial robots deployed globally, they currently power more than 45 robotics and embodied-AI partners, establishing themselves as a key industry execution layer.

## Future Industry Impact

By focusing on the physical execution layer, Tianji Intelligence addresses the gap between simulation and reality. Their hardware, supported by significant funding from Hillhouse and Ant Group, enables robots to handle contact-heavy tasks. As AI models advance, this force-coupled data will be critical for robots to perform complex, adaptive human-like interactions.

## Sources covering this story

- https://www.digitaltrends.com/brc/gento-debuts-a-21-dof-hand-at-corl-2026-as-nvidia-launches-gr00t-2-with-tianji-marvin-pro/ — 

---

_Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._
