Bangkok, Thailand — The Faculty of Medicine Siriraj Hospital, Mahidol University, in collaboration with SCB 10X and its AI research unit Typhoon, has announced a significant advancement in medical artificial intelligence with the development of Typhoon-Si-Med-Thinking-4B, a compact medical reasoning AI model designed to support clinical diagnosis.
As global AI development continues to prioritize ever larger models, this collaboration presents an alternative direction: building specialized, efficient AI capable of delivering competitive performance with lower computational cost. Typhoon-Si-Med-Thinking-4B is a 4-billion-parameter model that can operate on a single consumer-grade GPU while demonstrating performance comparable to much larger international AI models.
Designed specifically for medical reasoning, the model reflects real-world clinical workflows rather than producing a single definitive output. Typhoon-Si-Med-Thinking-4B is among the first medical AI models trained to generate ranked diagnostic lists, presenting multiple potential diagnoses ordered by likelihood. This approach mirrors the differential diagnosis process used by physicians in practice and represents a shift from conventional large language models toward reasoning-based AI systems.
By emphasizing clinical reasoning over final answers, the model helps reduce the risk of automation bias and supports physicians in making informed decisions. Rather than replacing clinical judgment, Typhoon-Si-Med-Thinking-4B is designed to function as a reasoning assistant, augmenting the diagnostic process with structured medical insight.
The project reflects a shared vision between Siriraj Hospital and SCB 10X that goes beyond AI adoption to AI creation, driven by cross-disciplinary research collaboration. Clinical expertise, medical data, and real healthcare challenges from Siriraj Hospital are integrated with advanced AI research and engineering capabilities from Typhoon (SCB 10X) and SiData+, Siriraj’s data innovation unit. This collaboration demonstrates how deep domain knowledge and advanced AI research can be combined to create practical and scalable medical technologies.
The research framework and findings are presented in the paper titled “On the Robustness of Answer Formats in Medical Reasoning Models” The project began in February and led to the release of a preprint paper alongside the signing of a Memorandum of Understanding (MoU) on December 12, 2025.
Typhoon-Si-Med-Thinking-4B is currently in the early research stage and has not yet been validated for clinical deployment. The primary objective is to share medical AI knowledge developed by Thai researchers and to enable further research and innovation that can benefit healthcare systems and society.
Further technical details are available at opentyphoon.ai. The project and its implications will also be presented at SiData+ Conference 2026, to be held on Friday, January 23, 2026, at Siriraj Hospital, Faculty of Medicine, Weikij Varanusorn Auditorium, 4th Floor, Atsadang Building. Registration is available at: https://www2.si.mahidol.ac.th/sirirajconference/event/sm/content/3528/TH/34276
