GigaChat MAX, the neural network developed by Sber, has completed a formal accreditation in General Medicine at Sechenov University’s accreditation center, a program aligned with the Russian Ministry of Health. The evaluation is designed to determine whether an AI system can support a practicing clinician in a core medical specialty, meeting the standards for medical knowledge, clinical reasoning, and patient safety. This milestone sits within Russia’s broader effort to explore artificial intelligence as a practical partner in healthcare, offering clinicians access to rapid, evidence-based guidance during patient encounters, triage decisions, and routine analyses. By demonstrating competence across essential medical topics and reasoning tasks, GigaChat MAX signals a growing willingness to test AI tools not only in theory but in the daily workflow of doctors. The achievement reveals a path toward integrating AI into mainstream medical practice with appropriate oversight and validation.
Sergei Zhdanov, director of the Sber Health Industry Center, described the accreditation as a meaningful step toward the practical use of artificial intelligence in healthcare. “GigaChat MAX continues to learn and get smarter to bring value to people. We will develop GigaChat MAX in individual specialties”, he said.
Two-stage testing was used to gauge both theoretical knowledge and applied clinical reasoning, reflecting real-world clinical demands. The first stage assessed a broad range of medical knowledge through a set of questions, while the second stage presented situational problems that mimic real patient scenarios to test decision-making under realistic constraints. This approach mirrors how clinicians consolidate knowledge with practical judgment in busy medical environments, ensuring that the AI can translate theory into meaningful guidance at the point of care.
During the first stage, the neural network answered 80 questions, correctly addressing a majority of them and surpassing the required pass threshold. It successfully answered 83 percent of the questions, exceeding the 70 percent pass mark. In the second stage, the model tackled situational problems and produced 20 correct answers, meeting and surpassing the minimum pass requirement of 17 correct responses.
Andrei Svistunov, the first vice-rector of Sechenov University, noted that such tools can increase the accuracy of clinicians’ decisions. “GigaChat MAX will save time and reduce routine workload by helping the doctor make decisions”, he said.
To place this achievement in a broader context, February 2024 saw GigaChat among the first AI systems worldwide to pass the final exam for general practitioner qualification. Medical graduates must undergo accreditation that tests deeper knowledge, and to prepare for this, GigaChat MAX has been trained by hundreds of doctors using a substantial corpus of medical training material. This training helps ensure that the model can engage with real clinical material in a way that supports human doctors without replacing their expertise.
In North America, health systems are watching such advances with close interest as regulators and clinicians weigh how AI-driven decision support might fit within licensing, privacy protections, and patient safety standards. The potential benefits include faster, more consistent guidance across diverse settings and the ability to alleviate routine workloads, but success hinges on clear oversight, robust validation, and ongoing monitoring to maintain patient trust and safety.