Leju deployed 15 Kuavo humanoid robots at an automotive-parts factory in Zhejiang, China, where they were tasked with identifying, picking up, moving and placing small metal components at a production station. The trial put the machines on a live line for 1,000 hours, according to the report.
Each robot cycle took under 40 seconds, compared with an average of 15 seconds for a human worker—about 2.7 times as long. The system recorded a 98% success rate while handling 60,000 components, implying roughly 2% of operations failed. The report says this error level requires close monitoring and intervention to avoid added costs. It also notes that the robots could adjust their movements as component positions changed, without manual input.
Leju described the deployment as a real-world setting for developing its embodied-AI systems, not simply a hardware test. The factory generated more than 40,000 operational data records per month for AI training, the article says. While the robots remain behind people in speed and still need to meet demanding production standards, the report frames their extended factory use and data collection as evidence of rapid experimentation in China’s automation sector. It argues that improving speed and cost could eventually increase pressure on manual labor, though the trial does not show that workers have already been replaced.