ZKH and Intel Launch LingLong Insight 1.0, an Edge AI Model for Industrial Sites
ZKH Group's AI subsidiary LingLong Intelligence has launched LingLong Insight 1.0, a lightweight industrial multimodal AI model built with Intel and designed to run at the edge. According to ZKH, the model helps workers verify delivered industrial supplies directly at the workstation.
- LingLong Intelligence, ZKH's AI subsidiary, launched LingLong Insight 1.0 with Intel at Intel Connection 2026 in Suzhou.
- The 4-billion-parameter multimodal model verifies delivered industrial supplies using camera images and delivery documents.
- It runs locally on LingLong Agent Box with Intel Core Ultra processors, reducing latency and cloud reliance, according to ZKH.
- ZKH cites more than 27 million product records and approximately 4 billion product parameters from nearly 30 years in industrial supplies.
- Planned extensions include warehouse checks, production-line material verification, and equipment operation and maintenance.
A model aimed at goods-receipt errors
The launch took place at Intel Connection 2026 in Suzhou, China. ZKH Group Limited (NYSE: ZKH) describes itself as a leading maintenance, repair, and operations (MRO) procurement service platform in China. LingLong Insight 1.0 was developed by its AI subsidiary, LingLong Intelligence, and is powered by Intel's edge computing technology.
The company points to a persistent problem: the wide variety of industrial supplies makes accurate checking of incoming goods difficult. Two bearings can look almost identical yet differ in a single specification that decides whether they fit a particular machine. Workers must tell such items apart and compare them with delivery documents, a demanding process that is prone to error. According to ZKH, an undetected mismatch can let the wrong part into inventory, delay critical maintenance, or disrupt production.
How the verification workflow works
LingLong Insight 1.0 combines image and text understanding with industrial knowledge to separate similar-looking supplies and check their specifications. The model supports a visual verification workflow:
- Industrial cameras capture images of delivery documents, packaging and labels.
- The model identifies the items.
- It checks their specifications against the delivery documentation.
- It flags discrepancies alongside the relevant images, so workers can catch potential mismatches before confirming receipt.
The model has 4 billion parameters, a size ZKH says is intended to balance accuracy and processing efficiency for edge deployment. It runs on LingLong Agent Box with Intel Core Ultra processors, which combine CPU, GPU, and NPU computing. Processing happens locally at the workstation, which the company says reduces latency and reliance on the cloud and supports responsive on-site verification.
Industrial data as the foundation
ZKH says its industrial data and domain expertise are central to the model. Drawing on nearly 30 years of experience in industrial supplies, the company has accumulated more than 27 million product records and approximately 4 billion product parameters. By standardizing this data and connecting it with expert knowledge, LingLong Intelligence aims to produce structured, AI-ready knowledge for scalable industrial AI applications.
"Industrial AI is only as good as the knowledge behind it," said Yang Liu, Chief Technology Officer of ZKH Group and CEO of LingLong Intelligence. He added that with Intel's edge computing technology, the company can "take that intelligence beyond the cloud and onto the factory floor, where every decision counts."
What comes next
The launch is part of a wider engagement between LingLong Intelligence and Intel. The two companies intend to deepen their collaboration on industrial AI and jointly explore global market opportunities, combining Intel's computing platforms and ecosystem resources with LingLong Intelligence's industrial data and model capabilities.
LingLong Intelligence also plans to extend the model beyond goods receipt to warehouse inbound and outbound checks, production-line material verification, and equipment operation and maintenance. For industrial users, the release points to AI checks that run locally on site rather than in the cloud, though the announcement does not give pricing, availability or accuracy figures.
