ByteDexter
20 000 USD
ByteDance Seed (ByteDance)
ByteDexter V2 is ByteDance Seed’s linkage-driven research hand with 21 DoF, palm-integrated motors and tactile fingertips. It powers the GR-Dexter project, where AI models learn bimanual, dexterous tasks from teleoperation and human demonstration data.
Hand Score
5
Specifications and details:
| Availability | Prototype |
|---|---|
| Degrees of freedom, hands | 21 |
| Strength [kg] | 5 |
| Weight [kg] | 0.97 |
| Size | 219 mm height × 108 mm width |
| Number of fingers | 5 |
| Tactile/Force Sensors | Yes: high-density piezoresistive tactile arrays on all five fingertips, covering the tip, pad and side surfaces to measure normal force |
| Motor tech | 16 electric motors integrated in the palm driving a linkage transmission; passive fingertip joints coupled by four-bar linkages |
| Main structural material | N/A |
| Manufacturer | ByteDance Seed (ByteDance) |
| Nationality | China |
| Website | https://seed.bytedance.com/en/ |
Description
ByteDexter is a dexterous five-finger robotic hand from ByteDance Seed, the AI research team at ByteDance, TikTok’s parent company. ByteDance built it to help train AI models that control robots with human-like hands. Most robot AI systems today still use simple grippers. In contrast, ByteDexter lets robots handle tools and everyday objects much as people do. The second version, ByteDexter V2, adds an extra thumb joint while making the whole hand smaller.
2026 Humanoid Robot Market Report
Interested in the ByteDexter? Download 198 pages of exclusive insight from global robotics experts — uncover funding trends, technology challenges, leading manufacturers, supply chain shifts, and surveys and forecasts on future humanoid applications.
humanoid expert
The hand uses rigid linkages to move its fingers, which makes it strong, durable and easy to maintain. In addition, all its motors sit inside the palm, so it works as a self-contained module. Touch sensors cover each fingertip, including the pad and sides, to detect pressure in fine detail. ByteDance pairs two of these hands with robot arms and controls them through VR headsets and data gloves. As a result, its robots have learned tasks such as vacuuming a table and serving bread with tongs.
Download the Humanoid Robot Market Report here
Website: https://byte-dexter.github.io/
