The Sberbank team announced a significant update to its GigaChat neural network model, presenting a new "Conduct Research" function at SPIEF. Now, the "neural network" can conduct in-depth, multi-stage research, independently analyzing data from various sources.
GigaChat provides detailed and structured answers to questions. Users do not need to formulate complex text queries. The developers assure that the logic of the model is easy to trace. For example, if a person requests an analysis of trends in the real estate market in Russia, GigaChat will be able to prepare a structured report. It will indicate market leaders, current trends, and development prospects.
The neural network can also perform full-fledged scientific research. For example, when requesting information about new directions in research on the impact of digital currencies of central banks on the banking system, GigaChat will study current publications and provide an overview with key points and links to sources.
The "Deep Research" function is available via the "Conduct Research" button in the web version of giga.chat, as well as in the mobile application for Android, which can be found in the RuStore and AppGallery stores.
Read more on the topic:
GigaChat Neural Network Takes Control of Sberbank Channels
Now on home
Start of deliveries scheduled for 2027
Over 51,000 new motorcycles were sold in Russia in 2025
The car will take at least a year to assemble
The application's audience has reached 20 million users
The model will be included in the list of cars for taxis, price - from 2.25 million rubles
All parking lots of the "Administrator of the Moscow Parking Space" are connected to the service
The cars will be supplied to the Moscow Transport Service Directorate
Deliveries to India may begin in 2028
The technology provides automated search for all types of defects in power units
The plane flew 500 km, accelerating to 425 km/h
The plant stated that the information about the termination of purchases for models 6 and 8 is not true
Scientists are using the "Ekran-M" installation