Sber’s neural network model, GigaChat, has achieved a significant milestone by becoming one of the first artificial intelligence entities globally to pass the rigorous examination of a higher medical institution in the field of “Medical Business.” This qualification is a prerequisite for obtaining the title of “medical doctor.” Similar to any medical student completing six courses according to the federal state educational standard, Sber’s AI underwent testing and provided responses during the examination, ultimately earning a final grade of 4. The examination panel comprised professors specializing in therapy, surgery, obstetrics, and gynecology from the Institute of Medical Education at the Almazov National Medical Center.
The standard oral examination consists of three situational tasks covering therapy, surgery, obstetrics, and gynecology, along with 3-5 related questions such as diagnosing conditions, devising treatment plans, and recommending additional examinations. Additionally, GigaChat underwent a written examination consisting of 100 questions, achieving a score of 82 percent, well above the passing threshold of 70 percent.
Sergey Zhdanov, Director of the Sberbank Health Industry Center, expressed enthusiasm for GigaChat’s rapid evolution across diverse domains. He extended gratitude to the Almazov Center staff for their pivotal role in overseeing the model’s training and validation. Zhdanov outlined future prospects for leveraging GigaChat as a foundation for developing doctor-patient assistants, enhancing healthcare delivery, and supporting clinicians in their daily practice, emphasizing the transformative potential of large language models in advancing human-centered healthcare.
The training of GigaChat involved a six-month process, utilizing a comprehensive dataset totaling 42 gigabytes of specialized information. This dataset included recommended teaching materials from Russian medical schools, monographs, methodological guidelines, scientific articles, and anonymized medical records. Despite its capabilities, the model is not a replacement for medical professionals; its recommendations must be validated by attending physicians.
Evgeny Shlyakhto, General Director of the Almazov National Medical Research Center of the Russian Ministry of Health and President of the Russian Society of Cardiology, praised the collaborative effort involving hundreds of faculty, researchers, residents, and students. He expressed satisfaction with the current results and highlighted plans to further develop application solutions for medical institutions, patients, and doctors based on GigaChat, commencing this year. Shlyakhto extended appreciation to Sberbank for entrusting the Almazov Center with such a significant and socially impactful project partnership.
