Practitioners and Participants Behind Industry Standards: Decoding China's Medical Large Model 'National Team'

Recently, the “Governance Specifications for Large Model Synthetic Services in the Healthcare Industry – Part 2: Algorithmic Models” (Standard No.: 2024-0028T-IHIA) was officially released. Previously, the Internet Healthcare Industry Alliance had issued several important standards, including the “Requirements for Security Management Capabilities of Large Models in the Healthcare Industry” (2023-0037-IHIA), the “Data Processing Requirements within the Governance Specifications for Large Model Synthetic Services” (2023-0036-IHIA), the “Technical Requirements for Hospital Scenarios and Patient Services” (2023-0035-IHIA, 2023-0034-IHIA), and the “Technical Application Requirements for Clinical Research and Traditional Chinese Medicine” (2024-0026T-IHIA, 2024-0027T-IHIA). These efforts have established a comprehensive standard system covering multiple dimensions, including security management, data governance, and clinical applications.

The algorithm model standards released herein were jointly led by the China Academy of Information and Communications Technology (CAICT) and Xunfei Healthcare Technology Co., Ltd., and developed in collaboration with more than 80 medical institutions, research institutes, and informatization enterprises, including West China Hospital of Sichuan University and Beijing University of Posts and Telecommunications. Focusing on capabilities in medical knowledge and linguistic patterns, the standards establish specific specifications across nine core technical areas—such as pre-trained models, knowledge fusion, and reasoning—and six major security requirements, including data security. These standards provide a clear technical pathway and a robust security assurance framework for industry development, marking a new phase of standardized and regulated growth for large medical models in China.

The unique nature of the healthcare industry dictates that any innovative application must adhere to principles of rigor and reliability. A diagnostic deviation or a medication error can potentially endanger a patient’s life. From the perspective of daily hospital operations, large language model (LLM) technology has permeated multiple stages—including physician consultations, clinical decision support, medical imaging analysis, medication guidance, and patient management—thereby reshaping traditional diagnosis and treatment paradigms.

Since 2016, with the successive influx of capital, various AI healthcare startups have emerged rapidly, enabling the industry to leapfrog from market education to large-scale application. According to Global Market Insights, the compound annual growth rate (CAGR) of the “AI + Healthcare” market is projected to exceed 29%, with its size expected to reach $70 billion by 2032. Behind the frenzy of this trillion-yuan blue-ocean market, however, the path to commercial implementation faces numerous challenges: How can the accuracy of models be ensured in professional medical scenarios? How can sensitive patient data be protected? How can the boundary between innovation and safety be balanced? These questions have become significant obstacles confronting the industry.

Particularly in the current era, where large model technology is triggering a new round of industry transformation, the control of safety boundaries has become especially crucial. In healthcare scenarios, large models must not only accurately comprehend professional knowledge but also address challenges such as patient data security and medical privacy protection. It is precisely in response to these industry pain points that establishing a scientific and rational standard system has become an urgent priority.

Looking back at the global development of AI in healthcare, standardization and normalization have become industry consensus. Establishing unified evaluation criteria, defining clear safety boundaries, and delineating responsibilities not only provide direction for medical institutions and enterprises but also drive the entire industry toward greater standardization and reliability. As emphasized by the standards released herein, in healthcare scenarios, AI must possess not only “capability” but also “principles.”

As the standard system for large medical models continues to improve, the healthcare industry is advancing with greater stability. In this standardization process, which is crucial to the future of the sector, a group of key players deeply rooted in AI healthcare are playing a pivotal role. Among them, as the sole corporate representative involved in the formulation of these standards, IFLY TEK has contributed unique value through its years of accumulated technological expertise and industry insights.

In the highly specialized and high-risk field of healthcare, the application of AI technology has always faced a fundamental challenge: how to ensure that artificial intelligence truly serves clinical practice while maintaining safety and quality?

As the core strategic initiative of iFLYTEK in the healthcare sector, IFLY TEK has demonstrated its confidence in solving industry challenges through robust technological innovation. As early as 2017, its Intelligent Doctor Assistant passed the National Medical Licensing Examination with a score surpassing 96.3% of human candidates, becoming the only artificial intelligence robot globally to achieve this feat and laying a solid technical foundation for IFLY TEK in the field of AI healthcare. With the advent of the large language model era, IFLY TEK has accelerated its innovation: in October 2023, the iFLYTEK Spark Medical Large Model was officially launched; just one year later, in October 2024, version 2.0 was released, outperforming GPT-4 Turbo in six major medical NLP tasks, including expert-level medical knowledge graph question answering, and successfully deployed in over 300 clinical scenarios. Also in October 2024, the company unveiled the iFLYTEK Spark Medical Imaging Large Model and initiated an open collaboration plan for full-spectrum, multi-modal medical imaging large models, further enhancing its technological landscape and application scenarios. In January 2025, iFLYTEK made a significant move by launching iFLYTEK Spark X1, the only deep-reasoning large model on a fully domestic computing power platform, elevating the model’s clinical reasoning capabilities to new heights. Following this breakthrough, IFLY TEK announced that an upgraded medical large model based on Spark X1 would be launched in the first half of this year, continuing to lead industry innovation.

Notably, the iFlytek Spark Medical Large Language Model consistently adheres to a fully domestic technological roadmap, with training completed on domestically produced computing platforms. In the specialized field of healthcare, data security and patient privacy protection are non-negotiable. IFLY TEK’s fully domestic technological approach not only ensures autonomous control over core technologies but also establishes a robust security barrier—from underlying architecture to practical application—for the compliant use of medical data and the protection of patient privacy, thereby laying a solid foundation for the rapid development of medical artificial intelligence.

The value of technological innovation must ultimately be validated through clinical practice. IFLY TEK deeply understands this principle and ensures that technological innovations consistently address critical clinical pain points by building a comprehensive “industry-academia-research-clinical” ecosystem. Through in-depth collaborations with the Chinese Medical Association Publishing House and the Scientific and Technical Literature Press, as well as by establishing its own high-caliber team of medical experts, IFLY TEK has constructed a multi-dimensional professional knowledge graph encompassing disease knowledge bases, symptom and sign databases, and laboratory and diagnostic test repositories. This provides reliable knowledge support for AI-driven clinical decision-making. At the institutional level, strategic partnerships have been established with over 40 top-100 hospitals, including Peking Union Medical College Hospital, and seven top-10 hospitals, with a focused deployment in key specialties such as cardiology, neurology, and radiology to promote the deep application of AI in specialized medical fields. Meanwhile, by co-establishing the Joint Research Center for Natural Language Processing Technology with Tsinghua University, setting up multiple joint laboratories with the University of Science and Technology of China in areas such as AI for healthcare, and participating in medical research under the State Key Laboratory of Cognitive Intelligence, IFLY TEK continuously strengthens the academic foundation of its technological innovations, thereby forming a complete innovation chain.

These technological accumulations and ecosystem advantages have been translated into significant value in clinical practice. Taking the “Huaxi Hongyi” large language model, developed in collaboration with West China Hospital, as an example, it has increased the accuracy of connotation quality control for complex medical records to 90%. The cardiac ultrasound diagnostic decision support system, jointly created with Beijing Anzhen Hospital, has injected new AI momentum into the diagnosis and treatment of cardiovascular diseases. Currently, IFLY TEK’s services cover more than 60,000 medical institutions across over 30 provinces in China, having provided more than 877 million AI-assisted diagnostic suggestions cumulatively, and firmly securing the top position in the grassroots Clinical Decision Support System (CDSS) market with a share exceeding 60%. In terms of patient services, its post-diagnosis management platform has served over 260,000 patients, enhancing the continuity of medical care through intelligent follow-ups and health management.

From the perspective of the development trajectory of large medical models, competition among leading enterprises is extending from technological innovation to the breadth and depth of industrial practice. In this context, iFLYTEK Healthcare has demonstrated solid technical accumulation, as evidenced by prestigious honors such as the First Prize of the National Science and Technology Progress Award. According to Frost & Sullivan data, iFLYTEK Healthcare ranks first in revenue scale within China’s medical artificial intelligence industry, highlighting its leading advantage in commercial implementation.

An examination of iFlytek Healthcare’s development trajectory reveals that in the highly specialized and high-stakes field of healthcare, realizing the true clinical value of AI technology requires companies to simultaneously advance on three fronts: technological innovation, industrial practice, and ecosystem building. By leveraging its independently developed large healthcare models, iFlytek Healthcare has accumulated extensive practical experience across multi-tiered medical scenarios, ranging from primary care clinics to top-tier hospitals. This profound understanding of healthcare workflows has made it an indispensable participant in the formulation of industry standards. As the core healthcare initiative of iFlytek, a national champion in artificial intelligence, iFlytek Healthcare is defining the strategic positioning of China’s “national team” in large healthcare models through its deep cultivation of vertical domain applications.

On December 30, 2024, this enterprise, which has been deeply engaged in the AI healthcare sector for many years, successfully listed on the Hong Kong Stock Exchange, signifying significant recognition from the capital market. According to the latest research report by Haitong Securities, with the deep integration of AI technology into healthcare scenarios, IFLY TEK is expected to maintain its high-growth momentum. Its operating revenue is projected to surpass the RMB 1 billion mark between 2024 and 2026, with an average annual growth rate remaining above 23%. The capital market has also fully acknowledged its development prospects, assigning a dynamic price-to-sales (P/S) ratio of 14–15x for 2025, which corresponds to a reasonable market capitalization range of HKD 12.5–13.4 billion. These factors collectively underscore IFLY TEK’s core competitiveness in the AI healthcare track.

IFLY TEK’s exploratory practices demonstrate that large language models are reshaping traditional healthcare paradigms at an unprecedented pace. In this promising sector, leading internet companies such as Tencent and JD.com, along with numerous tech startups and established pharmaceutical firms, have entered the fray to actively explore applications within the medical industry. IDC projects that by 2025, the healthcare segment will account for one-fifth of the global artificial intelligence application market—a trend that presents both opportunities and challenges. Against this industry backdrop, the release of standardized guidelines has become particularly critical.

Taking the newly released algorithmic model specifications as an example, they precisely target the core needs in the development of large medical models. At the technical level, the standards focus on nine core capabilities, including pre-trained models, knowledge fusion and reasoning, and knowledge base construction and updating, providing clear guidance for models to master medical knowledge and linguistic patterns. This is akin to creating a detailed “roadmap” for the “growth” of large medical models, ensuring that technological innovation follows established rules while maintaining controllability in clinical applications.

In terms of security assurance, the standard establishes a comprehensive protection framework. Covering aspects from training data security and model security to application security, as well as system security, management security, and security audits, each requirement directly addresses the practical needs of healthcare scenarios. These specifications are not mere restrictions; rather, they establish clear safety baselines for industry development, enabling healthcare institutions and enterprises to advance AI adoption with greater confidence.

Notably, the standards place special emphasis on the development concept of human-AI collaboration. Rather than restricting AI development or blindly advocating for AI replacement, this approach charts an innovative path of “AI empowerment + physician leadership” for the healthcare industry. This philosophy aligns closely with the practical explorations of companies such as IFLY TEK, underscoring the forward-looking nature and practicality of the standards. Within this framework, AI technology will better serve clinical practice, becoming a crucial support for enhancing the quality of medical services.

With the implementation of these standards, the development of large medical models in China will enter a new phase characterized by greater standardization and order. It is foreseeable that, as the sole corporate representative involved in formulating the standards, IFLY TEK not only holds the “steering wheel” for industry development but is also poised to gain a first-mover advantage in the next round of competition, leveraging its deep understanding of the standards and extensive practical experience. Judging from the trajectory of industry development, driven by the dual forces of “standardization” and “capability,” companies like IFLY TEK that have deeply cultivated the healthcare sector are expected to pioneer successful commercialization cases in the era of large models, setting new benchmarks for the entire industry.