Consumer-facing embodied intelligence firm Knowin Intelligence announced it has recently closed a funding round worth several hundred million yuan, led by a JD.com-affiliated fund, with participation from Zhengxin Valley Capital, Nanshan Zhanxin Investment, and Walden International.

This marks the fifth financing round Knowin Intelligence has completed since its founding, bringing its cumulative fundraising to more than 1 billion yuan. Just 49 days earlier, on August 10, the company disclosed a 500 million yuan angel-plus-plus round led by Matrix Partners China.

Compared with the previous round, this round's investor list now includes a JD.com-linked fund. For a company developing robots for household scenarios, the entry of industrial capital has drawn greater attention to its consumer market, channel, and supply chain strategy.

The new funds will be directed toward three main areas: expanding training data to improve the generalization capability of its GLOW generative embodied large model across different tasks and home environments; continuing development of the robot hardware to advance engineering validation and mass production of its first product, KNOWIN-X1; and expanding its R&D team.

Founded in August 2025, Knowin Intelligence focuses on consumer-grade home robots. The company positions GLOW and KNOWIN-X1 as the robot's "brain" and "body" respectively, aiming to let robots adapt to different household spaces, objects, and usage habits through coordination between model and hardware.

Home robots have largely remained at the stage of single-task demonstrations. The problem Knowin Intelligence is currently trying to solve is enabling a robot to learn new tasks from a single human operation.

According to the company, the new version of GLOW can simultaneously understand the human operation process, the on-site environment, the robot's state, and execution history, translating a single demonstration into reusable skill information.

In live demonstrations shown by the company, the robot can transfer a watering task to different containers, and can also continue completing box-opening and storage tasks after changes in objects, positions, and operating methods. Its learning focus is no longer a fixed motion trajectory, but rather task goals, object relationships, and execution progress.

Knowin Intelligence also released results from multiple model evaluations. Under its disclosed testing conditions, GLOW achieved an average success rate of 62.2% across 18 simulated tasks on RoboDojo, and an average success rate of 86.7% on the LIBERO-Pro benchmark; its embodied scene understanding model KnowinBrain-1.5 scored 65.62 points on the 2D embodied question-answering leaderboard of Embodied Arena.

To advance both model and robot hardware development simultaneously, Knowin Intelligence has assembled an R&D team of more than 200 people, covering algorithms, software, and hardware engineering. The company recently brought in Chen Yingcong as a scientific advisor on multimodal models, and Deng Shiyuan as partner and head of infrastructure, strengthening its capabilities in visual generative models and computing infrastructure respectively.

Following its密集 fundraising, Knowin Intelligence is about to enter the product commercialization phase. According to plan, KNOWIN-X1 will be released in the first quarter of 2027 with mass production to follow. From now until product launch, the company's focus will shift from model performance and live demonstrations to production, cost control, and sustained operation in home environments.

Home scenarios are far more complex than laboratories, as robots must contend with constantly changing spatial layouts, object states, and user needs. Whether KNOWIN-X1 can achieve mass production on schedule and turn "learn-from-one-teach" into a stable, repeatable user experience will be the first product test Knowin Intelligence faces after completing five funding rounds.