GENTO Shows Off 21‑DoF Dexterous Hand as NVIDIA Presents GR00T 2 with TIANJI Marvin Pro at CoRL 2026

During the 2026 Conference on Robot Learning (CoRL), robotics companies GENTO and NVIDIA each spotlighted major strides in robot manipulation: GENTO introduced a 21‑degree‑of‑freedom (DoF) hand, while NVIDIA unveiled its second‑generation GR00T platform powered by the TIANJI Marvin Pro computing system.

The GENTO hand was demonstrated live and boasts a densely packed joint architecture that permits finger movements as nuanced as those of a human hand. Built with tendon‑driven actuation and a lightweight exoskeleton, the prototype can seize, rotate, and handle objects of diverse geometries, a capability researchers say may bridge the gap between lab‑scale prototypes and field‑ready manipulators.

NVIDIA took the same stage to roll out GR00T 2, an enhanced iteration of its General Robotics platform that incorporates the high‑performance TIANJI Marvin Pro board. The showcase featured the robot executing fast pick‑and‑place cycles, reorienting irregular parts, and adjusting its grip on the fly, all powered by on‑board AI accelerators and ultra‑low‑latency sensor processing.

Both releases highlight a growing fusion of sophisticated hardware and AI‑driven software within the robotics industry. While GENTO’s hand pushes the mechanical limits of manipulation, NVIDIA’s GR00T 2 leverages strong compute to turn sensory input into exact control signals, a pairing that could speed adoption in manufacturing, logistics and service sectors.

The new hardware builds on earlier milestones: GENTO’s former 15‑DoF hand and NVIDIA’s original GR00T system, each of which laid the groundwork for tighter tactile sensing and machine‑learning‑based control. By adding more joints and expanding computational bandwidth, the latest offerings aim to tackle persistent challenges such as handling fragile items and operating in unstructured environments.

Looking forward, both firms said they plan to broaden collaborations with academic laboratories and industrial partners to test the technologies on benchmark tasks and real‑world applications. As the robotics community continues to chase greater dexterity and smarter autonomy, the tools revealed at CoRL 2026 could become reference points for the next generation of adaptable robot manipulators.