From Lighthouse Factories to Industrial Brains: The Practice of the Chinese Solution of the KAIROS Industrial Large Model
At 2 p.m. on the afternoon of July 26, 2025, the WAIC (World Artificial Intelligence Conference), known as the largest of its kind in history, opened at the Shanghai World Expo Exhibition and Convention Center. With an exhibition area of 70,000 square meters and more than 800 exhibitors, it presented over 40 large - model products.
Last year, when people talked about the large - model revolution in artificial intelligence, the most frequently mentioned were still general large models like ChatGPT, Sora, or Gemini. However, this year, the large models in the industrial field have undoubtedly attracted more attention from the industry. Behind this transformation are the dual drivers of the urgent need for the intelligent transformation of the manufacturing industry and technological breakthroughs.
From the perspective of industry demand, the manufacturing industry is undergoing a profound transformation from digitalization to intelligentization. Enterprises are no longer satisfied with the display of single - point technologies but need AI solutions that can solve practical pain points.
The industrial large model precisely responds to this demand - by integrating multi - source data such as equipment data, process parameters, and sensor information, it constructs a full - process intelligent system covering R & D, production, and operation and maintenance.
Facing the challenges of the in - depth application of industrial AI, Chen Lucheng, the Chairman and General Manager of Kaos Internet of Things Technology Co., Ltd., told 36Kr, "The process mechanisms vary significantly across different industries. How to combine general AI capabilities with in - depth industry knowledge is the direction that Kaos has been continuously exploring."
At this conference themed "In the Intelligent Era, All Nations Work Together", Kaos presented its latest technological achievements in the fields of industrial large models, digital twins, and industrial agents, as well as its practices in enabling the digital and intelligent transformation of manufacturing enterprises, demonstrating a new ecosystem of the innovative integration of AI and the industrial Internet.
The Advancement of Industrial Large Models: A Qualitative Change from Single - Point Intelligence to 'Full - Domain Collaboration'
Walking into Kaos' exhibition area, from basic weather queries and exhibition area guides to professional interpretations of industrial knowledge in 15 industries and dissections of digital practice cases, the digital avatars of the Haier Brothers presented by 3D stereoscopic projection can always respond quickly in colloquial expressions and even adjust the depth of explanations according to the industry background of the questioner.
This kind of interactive experience, similar to chatting with an 'old master', has impressed many manufacturing industry practitioners visiting the exhibition. Such a natural and smooth interaction is not simply an AI dialogue technology but an intuitive presentation of the 'perceptible, conversational, and reusable' industrial knowledge.
What supports all this is the Tianzhi Industrial Large Model of Kaos. Kaos COSMOPlat has built a complete system of 'efficient data flow - in - depth model empowerment - precise execution by agents', transforming industrial data from static assets into dynamic productivity.
Its underlying support is the data highway constructed by the digital industrial operating system. Through the technological foundation of 'large - scale connection + big data', it realizes the full - domain connection between the device end and industrial clusters and the data integration of the upstream and downstream of the industrial chain.
At the exhibition, Kaos COSMOPlat built an AI technology system integrating the'sky, air, and ground', presenting the multi - level applications of industrial intelligence in a panoramic way, which became one of the most attractive highlights on - site.
'Sky': With the Tianzhi Industrial Large Model as the top - level technological support, it accumulates industrial knowledge and drives intelligent innovation;
'Air': With industrial agents and industrial software in industries such as home appliances and chemicals as the air - link, it can be flexibly adapted to various industrial scenarios;
'Ground': With intelligent terminals and industrial robots as the means of implementation, it truly applies AI capabilities to the front - line of production.
The full - link standardized design from collection standards, governance rules to circulation mechanisms endows industrial data with the natural characteristics of 'connectivity, usability, and vitality', providing a continuous supply of high - quality 'fuel' for industrial agents.
Looking from the booth at the Shanghai World Expo Exhibition and Convention Center to the broader industrial landscape, Kaos' practices are answering a core question - What is the ultimate value of industrial AI?
It is to enable every factory to efficiently acquire intelligent capabilities and every industrial scenario to achieve value enhancement.
To achieve this goal, as the 'nerve center', the Tianzhi Industrial Basic Large Model, relying on the multi - modal industrial data set accumulated at the base, has developed leading capabilities in industrial scenario analysis, knowledge answering, and logical reasoning. It can not only understand professional drawings and process standards but also conduct decision - making deductions based on data, providing core support for cognition and decision - making for industrial agents.
More importantly, when building the large model, Kaos did not confine itself to its own industries but is committed to abstracting the 'cross - industry migration ability'. It modularizes knowledge and constructs an agent development toolkit, allowing enterprises to reuse components according to their own scenarios.
Feng Xingzhi, the Deputy General Manager and CTO of Kaos, said at the AI + Ecosystem Forum that the Tianzhi Industrial Large Model of Kaos, empowered by AI + industrial Internet technology, has created a unique model of 'one industry, one model'. It has been implemented in more than 40 scenarios in 9 industries and is also the industrial large model with the most cases of empowering Fortune Global 500 companies.
This combination of 'industry knowledge accumulation + open model platform' may be the key to the real scalability of industrial intelligence in the future.
Overcoming the 'Tough Nuts': In - Depth Application Shows the Real Results
As Chen Lucheng said, "The needs of industrial enterprises are 'one meter wide and one hundred meters deep'. The differentiated needs of different industries and scenarios require us to deeply explore industry knowledge." The promotion of industrial large models faces many challenges, especially how to combine general AI capabilities with in - depth industry knowledge.
At the conference, Kaos also presented solutions for different industries. At the Kaos AI + Ecosystem Forum on July 27, many new achievements in AI + industrial Internet were unveiled. Among them, the Tianzhi·Petrochemical Large Model of Kaos is the world's first petrochemical large model worth hundreds of billions, covering the entire industrial chain, ecological chain, and value chain.
Based on the technological foundation of the Tianzhi Industrial Large Model and the platform's practical experience in the chemical industry, this large model covers the entire 'oil, gas, coal, chemical, and electricity' industry, integrates the entire production chain of 'procurement - production - logistics - sales - service', helps chemical enterprises shift from mass production to mass customization, and will cooperate with industry partners to build an open intelligent ecosystem, providing a more inclusive digital path for the petrochemical industry.
For example, the world's first petrochemical large model covering the entire industrial chain, jointly built by Kaos and Yanchang Petroleum, integrates the data of the entire 'oil, gas, coal, chemical, and electricity' industrial chain, supporting the precise decision - making of 38 agents such as the dynamometer card intelligent assistant.
Taking the dynamometer card intelligent assistant as an example, the agent can not only 'understand' the dynamometer cards and relevant data of industrial equipment such as deep - well pumps, complete chart drawing and analysis, but also accurately identify equipment failures and give disposal suggestions, improving the efficiency of fault diagnosis and disposal of pumping well equipment by more than 50%.
Also launched at the forum, the Tianzhi·Energy and Carbon Large Model of Kaos is the first large model with hundreds of billions of parameters in the field of energy and carbon, originating from the practices of sustainable lighthouse factories. Combining more than 30 years of energy and carbon management experience in smart energy, on the basis of the large model, Kaos COSMOPlat develops functions such as knowledge answering, data analysis, user early - warning, report generation, and intelligent terminals, empowering the entire process of user - side energy consumption management in an integrated mode and helping thousands of industries achieve green, low - carbon, and sustainable development.
By creating the 'Energy Xiaozhi' agent, the platform innovates the user interaction form. Managers can interact with it through natural language, realizing data aggregation, alarming, one - key work - order transfer, diagnosis, etc. while 'chatting', which helps enterprises upgrade their energy management models.
For example, through the Kaos solution, the intelligent air - compressor station transformation project of Xinjin Group achieved an energy - saving rate of 30% and a carbon reduction of 26,000 tons over the entire life cycle.
These cases prove that industrial large models are not only technological tools but also the core engines for reconstructing the value chain of the manufacturing industry. The above achievements have also promoted the improvement of the brand value and market recognition of Kaos COSMOPlat. It has ranked first in the brand value of the domestic industrial Internet industry for 7 consecutive years, with a brand value of 116.335 billion yuan in 2025, being the only brand worth over 100 billion in the industry.
In the New Era of Industrial Intelligence, When Models Become 'Infrastructure'
Kaos is not the only platform that has proposed industrial large models, but it is one of the platforms that have gone the farthest in terms of engineering capabilities, industry scenarios, and ecological governance. What it shows is not only how AI technology is embedded in the industry but also how AI reshapes the fundamental logic of industrial operation.
"The complexity of industrial data far exceeds that of the general domain. It is multi - source and heterogeneous, and has extremely high requirements for real - time performance and reliability," Chen Lucheng told 36Kr. Kaos has built an integrated solution from data collection, mining to governance through the industrial Internet platform, providing high - quality 'fuel' for industrial large models.
Kaos' strategic path is not only to make breakthroughs through technology but also to focus on promoting the full - scale implementation of industrial large models by building a sustainable and evolving intelligent operating system. This operating system integrates the industrial Internet platform, AI technology, big data governance, and agent applications, realizing the full - chain empowerment of industrial intelligence.
At the AI + Ecosystem Forum, Kaos jointly released the first white paper on the practice of industrial large models in the industrial manufacturing field with Gartner - "The Value - Leading Practice of Kaos Tianzhi Industrial Large Model". Through the interpretation and analysis of the core capabilities, industry applications, and typical cases of the Kaos Tianzhi Industrial Large Model, combined with Gartner's strategic insights on generative AI in the manufacturing industry, it provides a practical path reference for the construction and practice of industrial large models, helping AI become the 'new - quality productive force' available to thousands of industries.
Moreover, Kaos' frontier exploration in AI + industrial Internet has also won international recognition. At the forum, the international standard for the IEEE intelligent interaction engine was officially launched, upgrading Kaos' exploration of the industrial structure of large models + agents to an international standard and promoting the high - quality development of the global AI + industrial Internet.
Conclusion: The Pathfinder and Leader
Looking back at the development context of the Kaos Tianzhi Large Model, it is not difficult to find that its success did not happen overnight. It is the digital crystallization of Haier's 40 - year manufacturing experience, and the abstraction and upgrading of thousands of on - site experiments, expert summaries, and industry knowledge.
As implied by its name 'Tianzhi', it hopes to endow the industrial system with the intelligent self - organizing ability for the future. Looking forward to the future of intelligent manufacturing, enterprises will no longer be organizations with factories but self - operating systems driven by models and data operation.
In the future, industrial large models will no longer be auxiliary tools but will serve as the central nerves for cognition, execution, and feedback in enterprise production and manufacturing. "Kaos aims to be the pathfinder and leader in industrial intelligentization, contributing wisdom to the industrial transformation in China and even the world," Chen Lucheng concluded.
Through the triple - drive of data, scenarios, and people, as well as the construction of standards and ecosystems, Kaos is upgrading industrial large models from technological options to necessary infrastructure, providing a Chinese solution for the global industrial intelligent transformation.