China’s medical AI sector is turning toward large language models, with at least 18 medical models developed for uses that include diagnostics, telemedicine and decision support. The shift brings conversational systems into a field previously associated more closely with medical imaging, while leaving questions about how AI should fit into clinical work.
A fast-growing market draws investment
Chinese healthcare companies have been exploring how large language models might support new medical applications. VBData estimates that China’s medical AI sector will grow at an average annual rate of around 40% between 2020 and 2025, with the market exceeding RMB 30 billion or $4.1 billion by 2025.
The figures point to substantial commercial expectations, but market growth alone does not show whether these tools improve care. The models are being developed for a range of roles, from helping with diagnosis to supporting remote consultations and assisting medical professionals with decisions.
At a book launch, Airdoc founder Dalei Zhang described how early he expects medical applications to arrive. Asked whether healthcare would be among the first industries to adopt GPT, he replied, “Medical care is not the first batch, it is the zeroth batch.” The remark captures the ambition around the technology, rather than evidence that it is ready to take over clinical responsibilities.
MedGPT puts conversational AI in the spotlight
One of the prominent examples is MedGPT, developed by Medlinker. It took part in what the source describes as the world’s first double-blind study in which AI doctors and human doctors saw real patients at the same time.
In a competition involving more than 100 patients with conditions ranging from cardiovascular problems to kidney disease, MedGPT scored only 0.3 points lower than doctors from top tertiary hospitals. That result has attracted attention because it compares an AI system with experienced clinicians in patient consultations.
The reported score offers a snapshot of performance in that competition. It does not by itself establish how the model would perform across other settings or what role it should play in everyday care. The study’s significance lies in bringing AI and human doctors into a direct comparison with real patients, while leaving broader questions about clinical use open.
From medical imaging to chat-based tools
The growing interest in language-based systems signals a change in emphasis for medical AI in China. Earlier attention centered on medical imaging systems; now, chat-based applications are also part of the sector’s development.
Conversational interfaces make it possible to frame AI as a tool for discussion, consultation or decision support. The source describes potential applications across diagnostics, telemedicine and professional support, but does not suggest that one model can handle all medical needs. The range of proposed uses means that performance and usefulness must be considered in the context of the task.
This shift also changes the questions surrounding medical AI. Rather than focusing only on whether software can analyze images, healthcare workers and developers must consider how a language model responds to patient concerns and how its output relates to professional judgment.
Clinicians remain cautious about AI’s role
Medical professionals have taken a measured view of AI’s capabilities, according to Huxiu. Previous disappointments in medical AI, including IBM’s Watson project, show that technology on its own cannot resolve medical problems.
Deep learning has advanced, but the source emphasizes that human radiologists are still needed. It also reports that AI’s influence may create a need for more specialized radiologists who can adapt to AI-driven changes, a point made by Feiyue Wang, a researcher at the State Key Laboratory of Management and Control of Complex Systems, Chinese Academy of Sciences.
Together, these developments suggest a sector experimenting with more capable tools while retaining a central role for medical expertise. China’s large language model applications may broaden the ways AI supports care, but the source’s account presents their growth alongside a clear reminder: progress in the technology does not remove the need for people who understand medicine.