Seoul National University Hospital has developed what could be the first medical large language model in South Korea.
The project to develop the medical LLM began last year in March. The SNUH research team started by collecting 38 million clinical texts, including hospital admissions, outpatients, surgical, prescription, and nursing records, and de-identified and anonymised them to create a base corpus for model learning.Â
Then, at the start of 2025, it developed department-specific knowledge bases that were later integrated into the LLM through retrieval-augmented generation (RAG). These knowledge bases consist of local medical laws, paper abstracts, and treatment guidelines in Korean, as well as medical terminology standards and abbreviation dictionary. Â
The LLM was tested on questions from the Korean Medical Licensing Examination of the past three years and scored 86.2%, which surpassed the average takers’ score of 79.7%. SNUH also said the model has shown high translation performance, processing 50,000-word texts simultaneously.Â
The SNUH research team will verify the LLM’s performance and safety for a year before it can be applied to assist clinical work and research. They also plan to expand its application in various medical fields and further enhance the model’s medical data processing capability.Â
WHY IT MATTERS
According to SNUH, existing medical LLMs developed by tech giants, including Google’s Med-PaLM 2 and Microsoft’s LLaVA-Med, are limited to comprehending Korean medical texts, laws, and treatment guidelines. As these models are optimised in the Western context, these are found unreliable for Korean medical professionals. Hence, the hospital researchers began developing a Korean medical LLM to meet the needs of local clinicians.
THE LARGER TREND
Late last year, a startup backed by conglomerate SK Group claimed to introduce the first LLM-based search platform specific to healthcare in South Korea. PhynX Lab’s Cheiron assists pharmacists and pharmaceutical industry researchers in retrieving relevant information by picking up the user’s intent through natural language processing. It can also self-reverify information and quickly generate high-quality answers.
Korea University Anam Hospital is also reportedly developing its own LLM, which is expected to be piloted this year. Â
Asan Medical Center has also utilised an LLM to create its voice-based clinical scribe, which is now used across its 16 departments.Â
Meanwhile, SNUH is currently working on other AI projects to free clinicians from administrative work. This includes a multimodal AI that can automatically generate patient summaries, AI for enhancing insurance claim processing, and AI that curates the latest papers for researchers.Â
ON THE RECORD
“Through the development of this Korean-style medical LLM, we have opened a new chapter in medical innovation by maximising the work efficiency of medical staff and providing faster and more accurate medical services to patients,” said SNUH president and CEO Dr Kim Young-tae in a statement.Â
Seoul National University Hospital has developed what could be the first medical large language model in South Korea. The project to develop the medical LLM began last year in March. The SNUH research team started by collecting 38 million clinical texts, including hospital admissions, outpatients, surgical, prescription, and nursing records, and de-identified and anonymised them to create a base corpus for model learning. Then, at the start of 2025, it developed department-specific knowledge bases that were later integrated into the LLM through retrieval-augmented generation (RAG). These knowledge bases consist of local medical laws, paper abstracts, and treatment guidelines in Korean, as well as medical terminology standards and abbreviation dictionary.  The LLM was tested on questions from the Korean Medical Licensing Examination of the past three years and scored 86.2%, which surpassed the average takers’ score of 79.7%. SNUH also said the model has shown high translation performance, processing 50,000-word texts simultaneously. The SNUH research team will verify the LLM’s performance and safety for a year before it can be applied to assist clinical work and research. They also plan to expand its application in various medical fields and further enhance the model’s medical data processing capability. WHY IT MATTERS According to SNUH, existing medical LLMs developed by tech giants, including Google’s Med-PaLM 2 and Microsoft’s LLaVA-Med, are limited to comprehending Korean medical texts, laws, and treatment guidelines. As these models are optimised in the Western context, these are found unreliable for Korean medical professionals. Hence, the hospital researchers began developing a Korean medical LLM to meet the needs of local clinicians. THE LARGER TREND Late last year, a startup backed by conglomerate SK Group claimed to introduce the first LLM-based search platform specific to healthcare in South Korea. PhynX Lab’s Cheiron assists pharmacists and pharmaceutical industry researchers in retrieving relevant information by picking up the user’s intent through natural language processing. It can also self-reverify information and quickly generate high-quality answers. Korea University Anam Hospital is also reportedly developing its own LLM, which is expected to be piloted this year.  Asan Medical Center has also utilised an LLM to create its voice-based clinical scribe, which is now used across its 16 departments. Meanwhile, SNUH is currently working on other AI projects to free clinicians from administrative work. This includes a multimodal AI that can automatically generate patient summaries, AI for enhancing insurance claim processing, and AI that curates the latest papers for researchers. ON THE RECORD “Through the development of this Korean-style medical LLM, we have opened a new chapter in medical innovation by maximising the work efficiency of medical staff and providing faster and more accurate medical services to patients,” said SNUH president and CEO Dr Kim Young-tae in a statement.Â
