Sber's GigaChat neural network passes energy engineering exam

Sber announced that its artificial intelligence model, GigaChat, has achieved a new milestone. The neural network successfully passed academic certification at the National Research University MPEI in the fields of "Electric Power Engineering" and "Thermal Power Engineering," receiving a "good" grade. This is the first Russian-developed language model to pass exams in multiple engineering specialties simultaneously. This was reported by Ixbt.com .

The examination process was organized by Sber specialists, MPEI scientists, and experts from Rosseti. The university developed a special benchmark to evaluate large language models. In tests covering 24 subjects, GigaChat demonstrated its knowledge in engineering calculations, regulatory frameworks, and energy economics.

This achievement means the neural network could become a useful tool for engineers, designers, students, and energy industry professionals in the future. GigaChat has previously tested its knowledge in over 20 fields, including medicine, music theory, finance, and construction. Additionally, the model scored 10 percent higher than the passing grade required for the international financial analyst certification.