WeDoctor's large medical model ranks first in CMB evaluation, leading the way for medical AI

The medical AI track is heating up, with many players entering the field. With the intensification of industry competition, barriers such as technology and data are gradually becoming apparent.

According to the Zhitong Finance APP, recently, the domestic authoritative medical large-scale model evaluation platform CMB (Comprehensive Medical Benchmark in Chinese) updated the Chinese medical model evaluation (CMB-Exam) ranking on its official website. WeDoctor's large medical model ranked first with an average score of 91.71.

CMB, as a comprehensive and multi-level medical benchmark tailored for the Chinese medical environment, covers over 280,000 questions and complex case discussions, simulating real medical scenarios to comprehensively evaluate the medical knowledge and clinical consultation capabilities of large models. WeDoctor's achievement not only highlights its leading position in the medical AI field but also validates its key role in promoting the digitalization upgrade of traditional medical services.

WeDoctor has been strategically positioning itself in the medical AI track for many years. Leveraging its strong large model capabilities, it has built an AI-empowered closed-loop process for "medical, pharmaceutical, testing, health, and management," providing patients with full lifecycle health management from pre-diagnosis to post-diagnosis. By pioneering the commercialization path of 2H2C, it has explored new market space for the industry. According to media reports, WeDoctor's revenue continues to grow and is expected to reach nearly 10 billion RMB this year.

Image | WeDoctor's large medical model ranks first on the CMB medical test set

It is understood that WeDoctor's large medical model has significantly improved its adaptability and problem-solving skills in dialogue scenarios through various strategies such as alignment training, role dialogue supervised training, and reinforcement learning based on human feedback. With diverse real application scenarios such as AI hospitals and AI health communities, WeDoctor's large medical model has demonstrated outstanding performance in training data, effect evaluation, and practical applications, solidifying its leading position in the fiercely competitive medical AI track.

At the application level of large models, WeDoctor's independently developed "AI doctor, AI pharmacist, AI health manager, AI intelligent control" four intelligent bodies have been widely applied in many medical institutions, especially in health communities and AI hospitals, forming a full-process application loop covering pre-diagnosis, diagnosis, and post-diagnosis. In terms of data, WeDoctor's platform has accumulated over 400 million conversation data, 200 million medical records, and 200,000 doctor diagnosis data, providing a solid foundation for the deep learning and effective application of AI products.

By relying on models such as "AI health community" and AI hospitals, WeDoctor has pioneered the commercialization path of 2H2C, which is the first mature business model to land in the medical AI industry. Compared to the simple 2C path of the Internet hospital 1.0 era, 2H2C shows great potential in user scale and deep value-added services Using Weiyi's business in Tianjin as an example, Tianjin Weiyi Health Digital Hospital, as the leading unit, has collaborated with more than 266 community health service centers and over 2,000 sites and health rooms in the city to jointly establish the Tianjin Basic Digital Health Community. Through the unified deployment of the "Four Clouds" platform including cloud management, cloud services, cloud pharmacies, and cloud examinations, it provides Tianjin residents with integrated online and offline medical health services for "prevention, diagnosis, treatment, management, and health". Empowered by AI, the Tianjin Health Community serves over 18,000 people daily, with the basic service scale steadily growing.

Figure | Chronic Disease Management Center of Weiyi AI Health Community

AI-empowered chronic disease management has effectively improved management effectiveness. Currently, there are 238 chronic disease management centers established within the Tianjin Health Community, with nearly 200,000 chronic disease patients included in the chronic disease management center's file management. Combined with family doctor contract services, Tianjin is exploring a payment reform by implementing a per capita payment for family doctor contract services after disease-based payment.

The "AI Health Community" model has become an excellent scenario for the commercialization of 2H2C. Official data shows that over the past three years of operation of the Tianjin Health Community, 80.3% of contracted patients actively return to the grassroots level, outpatient volume of grassroots hospitals has increased by 23% to 50%, public satisfaction rate is as high as 97%, and the average monthly medical expenses for diabetic patients have decreased from 1643 yuan to 1255 yuan, a decrease of 23.6%.

Behind the commercialization of leading technology lies the pioneering advantage of Weiyi's strategic layout and the expanding medical AI ecosystem.

As early as 2017, Weiyi partnered with Zhejiang University to establish the Ruiyi Artificial Intelligence Research Center, whose research results have been applied on a large scale in Weiyi's business scenarios. Recently, the Medical Imaging Artificial Intelligence Key Laboratory jointly established by Weiyi, Zhejiang University, and the Second Affiliated Hospital of Zhejiang University School of Medicine has been recognized as a key laboratory in Zhejiang Province. According to insiders, Weiyi is leveraging the capabilities of this key laboratory to promote the construction of a medical large model evaluation platform and advance the formulation of industry standards for medical large models.

Figure | AI Applications in Weiyi's Artificial Intelligence Hospital Founder of WeDoctor, Liao Jieyuan, publicly stated years ago that AI will help millions of grassroots medical institutions in China "overtake on the bend". Today, as a pioneer who has been doing AI in the medical industry for the longest time and doing medical work in the AI industry for the longest time, Liao Jieyuan thinks and indeed does so.

It is worth noting that with the advantages of data-driven and ecological synergy, WeDoctor has built a WeDoctor MaaS (Model as a Service) product matrix based on medical large models as the technological foundation and extensive medical service scenarios as the application basis. This not only promotes the coordinated development of internal business but also opens up new opportunities for external cooperation.

With the support of AI, WeDoctor is accelerating the resolution of the "impossible triangle" dilemma faced by traditional medical care, establishing a unique business model and commercial model for itself. The development of medical large models requires high demands on technology research and development, real data, and application scenarios, thereby building high barriers to industry competition. It can be foreseen that as the circle of friends of WeDoctor's medical AI expands, more innovative service scenarios and business models will continue to emerge, injecting new impetus into the digital upgrade of the medical and health industry and accelerating the industry's intelligent transformation