On January 8, Leju Robotics and Alibaba Cloud signed a strategic partnership. The two will pursue full-stack AI collaboration spanning compute, AI platforms and foundation models, aiming to drive innovation and tighter coordination for humanoid robots across manufacturing, data collection, model training and multi-scenario applications.
The parties have already kicked off joint projects. They are hosting the first Embodied Intelligence Manipulation Challenge on Alibaba Cloud's Tianchi platform, focused on showroom service scenarios, and have integrated the Qwen foundation model — all toward building a more human-like intelligent guide robot.
Image source: Leju Robotics
The two will also deepen cooperation around humanoid-robot training grounds. In this field, Leju’s open-source large humanoid-robot dataset (LET) has surpassed 600,000 downloads since its release. On Hugging Face, the well-known global open-source community, it has ranked among the overall top 10 out of more than 22,000 datasets sitewide, and topped the embodied-intelligence dataset leaderboard.
Leju Robotics' technical groundwork in humanoid robots underpins the tie-up. Its open-sourced LET dataset, now widely used, has become one of the sector’s key data infrastructures. Next, Leju plans to fuse its multi-scenario know-how with Alibaba Cloud’s powerful AI foundation, jointly exploring cutting-edge integration across the humanoid "brain (decision-making and cognition), body (motion and control) and cerebellum (real-time response and coordination)." The goal is to deliver mature, standardized solutions at scale for industrial, research, commercial-service and household settings — accelerating the large-scale rollout of embodied intelligence.
Alibaba Cloud's core strengths are another pillar of the partnership. As a global leader in cloud computing and AI, it will provide secure, reliable and accessible full-stack AI infrastructure and services — spanning high-performance computing, advanced AI development platforms and the Qwen family of large models — to support the humanoid value chain from R&D to broad deployment.
