Zhitong Finance App learned that Paradigm Intelligent Technology Group Co., Ltd. (hereinafter referred to as “Paradigm” or “Group”, stock code: 06682) announced its business performance for the first three quarters (“within the reporting period” or “during the period”) ending September 30, 2026.
In the first three quarters of 2026, the Group's revenue growth accelerated across the board, and operating profit increased sharply. The Group achieved total revenue of RMB 6.747 billion (same below), an increase of 53.3% year on year; realized gross profit of about 2.26 billion yuan, an increase of 39.4% year on year. The growth rate jumped significantly from the same period last year. The consolidated gross margin was approximately 33.5%, a steady increase over the first half of the year. The demand from enterprises to improve the efficiency of AI resource management continues to rise, and big model reasoning is being implemented in more business scenarios, driving the Group's revenue growth to accelerate.
While maintaining business growth, the Group's cost structure has been further optimized. The fee rate for the first three quarters of 2026 decreased by more than two percentage points compared to the first half of the year. Operating cash flow has increased significantly, and repayment cycles have been significantly shortened. The Group's overall financial leverage remained low. RMB borrowing costs fell to 2.25 to 2.80% in the third quarter, and the financial situation remained stable.
Dai Wenyuan, founder, chairman and CEO of Paradigm, said, “In the first three quarters, the Group's revenue accelerated across the board, operating profit increased dramatically, and profitability further improved. As far as the paradigm is concerned, high-quality growth must not only expand in scale, but also settle into sustainable profitability. Enterprise demand for AI is changing: customers not only need stronger models, but also need to better manage computing power and models so that AI can operate efficiently and reliably in actual business. As AI enters more industries, the key to the next stage is to make the infrastructure that supports the development and operation of AI more solid. Over the past 12 years, we've worked hard to make AI more people use; in the future, we want more people to participate in improving AI, so that better AI can truly improve people's lives. This is Paradigm's new vision on its 12th anniversary — 'Everyone for AI'.”
Strong growth in AI platform business, steady increase in profitability
In the first three quarters of 2026, the AI Platform business achieved revenue of 5.392 billion yuan, an increase of 32.0% over the previous year. In the third quarter, the gross margin of the business increased steadily compared to the first half of the year. As of September 30, 2026, the AI Platform business had on-hand orders of about 9 billion yuan, achieving significant year-on-year and month-on-month growth.
The AI Platform business is the Group's main revenue pillar and uses an efficient asset-light operating model. The business grew rapidly while maintaining stable gross margins, further improving profitability, and continuing to expand its market share. As customers place increasing emphasis on AI sovereignty, data security, and internal control and governance, the Group remains confident about the long-term growth prospects of this business.
Around these requirements, the paradigm continues to consolidate the technical barriers of the platform. In terms of computing power resource management, heterogeneous computing power scheduling is achieved through technologies such as Hami vGPU; the recently launched ModelSphere further covers model adaptation, inference performance optimization, and token management and scheduling to help customers improve the efficiency of using existing resources and reduce the deployment and operating costs of AI applications. With the continuous improvement of product capabilities, customer stickiness continues to increase.
The API business has exploded, and gross margin has further increased
In the first three quarters of 2026, the API business achieved revenue of 1,014 billion yuan, an explosive increase of 1,556% over the previous year. Following a year-on-year increase in gross margin in the first half of 2026, the gross margin for the third quarter continued to rise significantly compared to the first half of the year.
Customer token demand continues to grow, and the Group is simultaneously focusing on the operational efficiency and resource use efficiency of the API business. By the end of the third quarter, the usable computing power of the API business reached 40,000, a significant increase of 42.9% compared to the end of the first half of the year.
Against the backdrop that customer demand for tokens continues to be strong and the financial costs of the API business are below the industry average, the Group maintains confidence in the continued growth of the business and further improvement in profit margins.
Looking at the business mix, the AI Platform business uses an asset-light model to achieve high-quality growth, while the API business undertakes rapidly growing model reasoning requirements. The two major businesses complement each other at the business and financial levels, laying a solid foundation for the Group to continue to unleash strong profitability and achieve leapfrog development.
Product and technology progress: improving enterprise AI resource management and expanding model layout
The Group continues to promote core product iteration and technical capacity building around the key needs of enterprise AI applications.
In September 2026, Paradigm was selected for the creation list of the first batch of national emerging industry development demonstration base demonstration enterprises announced by the Ministry of Industry and Information Technology, and joined the lineup of only 18 industry-leading enterprises in the country. In the “China AI Computing Power Management Platform Technical Capability Assessment, 2026” report released by IDC in late September, Paradigm ranked first in the overall rating based on heterogeneous computing power management, fine-grained virtualization, and deep accumulation of computing power services for intelligent devices. It is also the only manufacturer that received a perfect score in the four dimensions of computing power resource management, computing power operation and maintenance, technical architecture, and service delivery.
In terms of model development, Paradigm has launched the Phai series of pedestal models, including the general language model Phai-A1 and the video generation model Phai-M1, and plans to release weights for the two models in the near future. Phai-A1 targets application scenarios such as code development and intelligent assistants, and focuses on improving reasoning efficiency and long context processing capabilities; Phai-M1 focuses on live action film and AI skit creation, focusing on improving image quality and simultaneous audio and video performance.
In terms of domestic computing power, the paradigm completed Phai-A1's full-process training from post-training to reinforcement learning based on leading domestic computing power platforms, opening up deep collaboration between model training and domestic computing power platforms. This also makes Paradigm one of the few domestic enterprises that can complete the entire process of training from post-model training to enhanced learning based on the country's computing power platform, further demonstrating Paradigm's comprehensive strength in self-developed models, domestic computing power collaboration, and enterprise-grade Al technology implementation.
Vision: Moving from “AI for Everyone” to “Everyone for AI”
On the occasion of its 12th anniversary, Paradigm proposed a new vision, “Everyone for AI.” Over the past 12 years, the paradigm has adopted the vision of “AI for Everyone” to push AI into more industries and serve more people. With the rapid spread of large models and agents, AI applications are becoming more and more widespread, and the infrastructure that supports its development and operation is becoming more important.
Facing the future, Paradigm will continue to deeply cultivate AI infrastructure, focus on solving key bottlenecks that limit AI development and operation, and support the implementation of more AI applications through continuous infrastructure improvements. From “let more people use AI” to “make AI better”, Paradigm hopes to work with all parties to provide a more solid foundation for the development of AI, so that more efficient and trustworthy AI can better serve people's lives.