Qianjue Robotics, a tactile-intelligence company, has closed two consecutive rounds worth several hundred million yuan. The latest included BlueRun Ventures, CICC-affiliated funds and Jingya Capital. An earlier round drew top embodied-AI industry players, Tianji Capital and Jide Electric, and its shareholder list now spans Hillhouse, Yuanhe Origin and Futon Capital plus industrial leaders Li Auto, AgiBot and Jide.
The proceeds fund visuotactile sensor iteration and scaled delivery, real interaction-data collection, tactile simulation and dataset building, and continued work on tactile embodied models and tactile world models for robots, industry and consumer scenes.
Founded in May 2024, Qianjue has reached 10,000-unit-scale production of self-developed visuotactile sensors and serves more than 300 paying clients across robotics, manufacturing, consumer brands and research institutes. In the embodied-AI supply chain, touch is becoming a critical sensing add-on.
Why touch, not just sight
Vision helps a robot perceive and identify objects, but at grasping, assembly and inspection, slip tendency, contact force and surface-shape change cannot be read from vision alone. Touch is the sense that lets a robot understand physical contact and perform fine manipulation. Policy is leaning in too: Shanghai’s embodied-intelligence plan lists multi-modal sensing, and Shenzhen’s revised 2026 action plan names visual-tactile-language-action models as a priority.
The gap between needing touch and shipping it at scale is an engineering cliff. A high-precision sensor prototype is only a start. real deployment tests long-term use, batch delivery and system fit. Qianjue’s disclosed extreme-condition results show 97 per cent cross-unit and repeat consistency, a 5 million-cycle contact life, IP68 water resistance and a 240-hour continuous underwater test, using a three-colour-light visuotactile scheme that outputs multi-dimensional force, kinesthesia, slip and 3D shape.
With hardware scaling, the next step is turning interaction data into robot perception and decisions. Qianjue built the XTac UMI G1 wearable collector for human tactile data, the TacVerse real-touch dataset, and the X-TouchMind V1 vision-touch-language-action model that bakes touch into contact understanding, reasoning and action, while pushing a tactile world model that predicts physical-interaction outcomes.
The loop runs end to end: front-end hardware sets input quality, the data system stores real interaction experience, the model learns understanding and decisions, and new needs from real tasks feed back to hardware and data engineering. Beyond humanoids, Qianjue works deeply with AgiBot and is in batch application, integrating sensors directly into robot systems for dexterous-hand operations.
Editor’s note: This is an adapted translation of the original 36Kr report. It has been trimmed and restructured for readability for an international business audience. The full original (in Chinese) is at https://36kr.com/p/3989809262983942.