The GigaChat initiative, undertaken to train Sber’s large language model, stands as a pioneering effort of the Center for Artificial Intelligence in Medicine, which launched alongside Sber in mid-2023. This milestone was highlighted by Dmitry Kurapeev, Deputy Director General for Information Technologies and Project Management at the Almazov National Medical Research Center. He noted that the Almazov Center has cultivated unique expertise in generative artificial intelligence, a field drawing global interest and attention.
Kurapeev extended gratitude to the entire GigaChat team for their professional and enjoyable collaboration on the project, underscoring the collaborative spirit that propelled the work forward.
Elena Parmon, Director of the Almazov Central Medical Education Institute, pointed out that artificial intelligence certification was achieved for a medical university graduate for the first time, marking a historic convergence of medicine and advanced AI training.
Parmon expressed appreciation for everyone who prepared such an unconventional learner, including educators and technical staff who imparted both foundational and cutting-edge medical knowledge. She also thanked her family and Sber for supporting the initiative to teach artificial intelligence in medicine.
She emphasized that gathering knowledge, performing analysis, and synthesizing insights represent a substantial, ongoing effort. While a solid start has been made, the path ahead remains long and full of potential, she observed.
In the medical profession, doctors continually expand their expertise over a lifetime. The crucial elements are genuine care for patients and the ability to communicate effectively. With this perspective, the GigaChat project is being repositioned as a future assistant to both physicians and patients on the patient’s side of care.»
GigaChat, the neural network developed by Sberbank, recently completed an examination at the level of a general medical education program, aligning with the qualifications required for a medical degree. This achievement positions the model among the first in the world to reach such a milestone.
According to official disclosures, the AI demonstrated competence by answering questions at the level expected of a medical school graduate, earning a score that surpassed the passing threshold. The evaluation was conducted by a panel of faculty members specializing in internal medicine, surgery, and obstetrics-gynecology from the Almazov National Medical Research Institute for Medical Education. The system also completed a broader 100-question assessment with a solid score, exceeding the required benchmark.