Shanghai's First Industrial Vertical Large Model — LingShu "Baigong" Passes Cyberspace Administration Filing | Linear Capital Portfolio Company

On February 22, the Shanghai Cyberspace Administration released its latest batch of registered generative AI services. BaiGong Large Model, developed by Shanghai Lingyi AI Technology under LingShu Intelligent, successfully passed registration (Registration No.: Shanghai-BaiGong-202502050034), becoming the first industrial vertical large model to be registered in Shanghai.

Newly Registered Generative AI Services in Shanghai

(February 22, 2025)

Among All Registered Large Models Nationwide,

Industrial Vertical Models Account for Just 2%

Since large model technology began driving a new wave of intelligent transformation across industries in 2023, a flood of domain-specific models and applications has emerged both domestically and abroad. That same year, the Cyberspace Administration of China launched its generative AI registration process to review and assess the compliance, professional capabilities, and other core competencies of large model products and services.

To date, among tens of thousands of large model applications and services nationwide, only about 300 models have passed the Cyberspace Administration's evaluation. These have undergone rigorous assessment across multiple dimensions including specialized scenarios, corpus safety, content generation risk control, and user privacy protection, ensuring that generated content strictly meets the standards of Generative Artificial Intelligence Service Security Basic Requirements (TC260-003).

• Technology and Security Testing: Must build test question banks covering 31 categories of risk, achieve an illegal content refusal rate exceeding 95%, and pass technical reviews of multimodal data alignment and real-time inference;

• Data Compliance Governance: Corpora must pass dual verification through manual spot-checks and intelligent filtering, with a pass rate exceeding 98% and strict avoidance of overseas data source risks;

• Cross-Department Coordination: Goes through local preliminary review, central final review, and consultation with six major ministries, with the full process taking 3-4 months and a pass rate of less than 30% for projects submitted in 2024 alone.

Even under these already stringent requirements, industrial vertical large models make up only about 2% of all registered models across industries and domains — a tiny fraction relative to the vast application market for industrial large models. Only sector-leading enterprises that both deeply integrate AI + industrial application value and maintain strict standards for model development can accomplish this feat.

What Makes Industrial Vertical Models Valuable?

Only Specialization Enables Deep Application in Industry

While general-purpose large models remain stuck in parameter races, industrial vertical models have quietly sparked a revolution in manufacturing applications.

As DeepSeek swept through every industry, numerous manufacturing enterprises discovered that general-purpose large models, despite being "broad," were "not specialized" — still unable to solve their complex, professional real-world business problems.

By contrast, LingShu BaiGong Industrial Vertical Large Model, through deep integration of industry know-how and multimodal data, can truly "speak industrial language and execute industrial tasks," demonstrating unique advantages:

• Specialized Scenario Adaptability: Focuses on core industrial segments including design R&D, manufacturing operations, and materials management. For example, dynamic parameter adjustment in collaborative manufacturing of large equipment, and professional ship drawing comprehension and application — complexities that general-purpose models struggle to handle;

• Highly Trustworthy Output: Through industrial mechanism modeling and multimodal fusion, avoids the "industrial hallucinations" caused by probabilistic generation in general-purpose models;

• Compute Resource Efficiency: Adopts an "industrial base model + domain fine-tuning" architecture, achieving compute utilization rates up to 97.6% and reducing training costs by over 40% compared to general-purpose models. This significantly lowers application costs for manufacturing enterprises, truly solving the "last mile problem" of industrial large model deployment.

LingShu Intelligent's BaiGong Large Model is deeply rooted in industrial domains represented by high-end equipment manufacturing, crystallizing over two decades of industrial management wisdom from the LingShu team and incorporating successful experience serving hundreds of clients. The model demonstrates outstanding professional capabilities in delivering precise, specialized intelligent services such as industrial drawing design, production operations management, and supply chain control. LingShu Intelligent has already partnered with leading enterprises including ZPMC (Shanghai Zhenhua Heavy Industries), China Electrical Equipment Group, and Shanghai Electric, covering over 40 specialized industrial scenarios.

Actively Advancing Shanghai's AI+ Initiative,

LingShu Serves as a Key Solution Provider

In Shanghai's key industries such as heavy equipment manufacturing, LingShu Intelligent's AI technology has quietly transformed traditional workflows. Take the ZPMC project: LingShu Intelligent developed a Multi-Agent system that coordinates complex issues across design, procurement, and production through natural language interaction, improving project adjustment efficiency by 50% and serving as an "intelligent assistant" for engineers.

Technology deployment has been accompanied by recognition from industry authorities. In 2024, the "ETO Manufacturing Delivery Intelligent Agent" jointly developed by LingShu Intelligent and ZPMC won the SAIL Star Top 10 award — the highest honor at the World Artificial Intelligence Conference — was selected as a demonstration case by the China Academy of Information and Communications Technology (CAICT), and received four consecutive front-page reports in Jiefang Daily, becoming a benchmark for industrial AI applications.

In January 2025, Shanghai grandly hosted the "AI+" Action Promotion Conference, which designated AI + manufacturing as one of six critical domains. As an application leader in industrial AI, LingShu Intelligent is committed to achieving qualitative leaps in corpus construction, vertical model development, platform building, and product deployment. The company will bring significant efficiency gains to more manufacturing enterprises, advance the realization of "AI + Industry" in Shanghai, and has successively provided industrial large model application deployment for multiple centrally administered state-owned enterprises that hold dominant positions in their industries.

LingShu Intelligent will continue to delve into large model technology, developing multiple hardware and software solutions tailored to market demands based on BaiGong Large Model, constantly exploring the optimal path for industrial large model development, and bringing customers more reliable and cost-effective choices, contributing to Shanghai's ambition to build an internationally influential hub for AI industry innovation and development.

About Linear Capital

Linear Capital is an early-stage investment institution focused on "frontier technology + industry," targeting frontier technologies represented by data intelligence, digital new infrastructure, next-generation robotics, and technological transformation of traditional sectors (such as biomedicine, materials, and energy), applied across vertical industries to substantially improve industrial efficiency, empower solutions to pain points, and complete industrial upgrading — achieving excess returns through substantial increases in industrial value. The firm currently manages ten funds with total AUM of approximately $2 billion. Our investment stage focuses primarily on leading angel to Series A rounds, with individual investments ranging from $1 million to $10 million (or RMB equivalent). To date, we have made early-stage investments in over 120 entrepreneurial teams including Horizon Robotics, Kujiale, Sensors Data, Tezign, Rokid, Guandata, and Agile Robots. The combined valuation of Linear Capital's portfolio companies is approximately $20 billion. In the near term, Linear Capital is working to become the best "Data Intelligence Technology Fund," and in the long term, gradually build itself into the most influential "Frontier Technology Application Fund."