Reportedly: Ant Group's Robbyant Open-Sources LingBot-Touch Tactile Model for Robot Hands
Reported by Pandaily · not yet confirmed by the company or a second independent outlet. We update this page when it is.
Ant Group's Robbyant open-sourced LingBot-Touch, a tactile foundation model unifying magnetic, visuotactile and piezoresistive sensing. The company reports 14 to 26 point gains for four action models on 26 real-robot tasks.
China context
- Original name
- 蚂蚁集团
- Outside China
- Open weights · pandaily.com
- Claims
- Company-reported; not yet independently evaluated
- For builders
- Developers outside China can download the LingBot-Touch weights and test the model on their own robot hands, but must verify the reported performance gains independently.
- For investors
- The open-source release may accelerate commoditization of tactile perception, reducing the moat of proprietary tactile models and shifting value to robot hardware and integration.
Robbyant, Ant Group's robotics arm, has open-sourced LingBot-Touch, a tactile foundation model designed for robot hands. The model unifies three sensing modalities: magnetic, visuotactile, and piezoresistive. According to the company, integrating LingBot-Touch improved performance of four action models by 14 to 26 points across 26 real-robot tasks. The open-source release makes the model weights publicly available.
LingBot-Touch is a tactile foundation model that processes inputs from magnetic, visuotactile, and piezoresistive sensors. The reported gains of 14 to 26 points on 26 real-robot tasks suggest the model provides a generalizable tactile representation that can be transferred to different action models. The open-source release allows external researchers to verify these claims and test the model on their own robotic setups.
Robot developers outside China can now access a tactile foundation model that may reduce the cost and time needed to train dexterous manipulation skills, potentially narrowing the gap with Chinese robotics labs that have invested heavily in tactile sensing.
For robotics companies, LingBot-Touch offers a pretrained tactile encoder that could lower development costs for dexterous manipulation. For Ant Group, open-sourcing the model may attract talent and ecosystem partners while positioning Robbyant as a leader in embodied AI.
A verifiable next signal is whether independent researchers reproduce the reported 14 to 26 point gains on the 26 real-robot tasks using the open-sourced weights. Another signal is adoption of LingBot-Touch in non-Chinese robotics projects or benchmarks.