At the SPIEF, they talked about the new features of GigaChat

The Sbera GigaChat neural network model has learned how to conduct in-depth research, including searching for information and providing a detailed answer. The "Conduct Research" function, designed for professional information analysis, can conduct high-quality multi-stage research: it independently analyzes data from various sources and provides detailed and structured answers to questions with links to sources. The updated functionality of the model was presented at the St. Petersburg International Economic Forum (SPIEF).

As explained by the experts of Sber, GigaChat has made a qualitative leap, moving to a fundamentally new level of user interaction. The intelligent agent orchestrator helps analyze incoming requests and forms teams of specialized AI agents to process them. The multi-agent neural network system allows you to simultaneously use many expert modules, providing a comprehensive analysis of the request and providing the most complete information in the responses.

The deep research mode allows specialists to save time by entrusting the AI model to routine work — collecting information, analyzing sources, and composing text. In this case, the user can follow the logic of the model when preparing the response.

The "Conduct research" function can help in preparing a qualitative analysis of trends in a particular market, analyze the products and strategies of competitors who operate in a similar business niche, as well as conduct a complete scientific study by examining relevant publications in the chosen field and providing an overview with key abstracts and links to sources.

The "Deep Research" feature is available via the "Do research" button in the web version of giga.chat. You can also test the new features of GigaChat in the mobile application of the same name for Android devices.