Sber’s GigaChat neural network model outperformed American research organization OpenAI’s ChatGPT-3.5 version in terms of the quality of responses in both English and Russian. In this respect reports economic information agency “Prime”.
As noted by Andrey Belevtsev, senior vice president of Sberbank and head of the Technology block, the bank’s specialists conducted the MMLU (Massive Multitask Language Understanding) exam. It included questions in 57 areas of knowledge, including mathematics, history, medicine, physics, world knowledge, and problem-solving skills.
According to Belevtsev, GigaChat also surpassed GPT in the quality of Russian responses.
“Sber launched GigaChat in test mode in April last year, and in early September the service became publicly available. It is used by more than 2.5 million people today. GigaChat can solve many intellectual or everyday tasks: maintaining a conversation, writing text or code, answering questions. And the inclusion of a Kandinsky model in the ensemble gives it the ability to create images,” said the head of Sberbank’s Technology block.
The company’s top executive also added that Sber continues to develop GigaChat for its business needs.
“Companies have access to the GigaChat API, a software interface that provides access to Sber’s neural network model. With its help, businesses can create their own solutions and optimize their internal processes. In addition, the service makes it easier to work with large texts, write articles in the desired style and format, write information helps search and prepare analyzes based on it,” said Belevtsev.