Sber’s GigaChat Adds Memory to Deliver More Personalized Responses
Updated AI assistant adapts to users and responds faster
Sber has updated its GigaChat AI assistant with a memory feature that allows it to remember users and deliver more personalized responses. The system can now tailor recommendations and interactions based on who is asking, while generating answers up to twice as fast.
The AI stores information such as a user’s activities, interests, preferences, hobbies, pets’ names, habits, and details about close contacts, based on what the user chooses to share. It automatically identifies relevant details, while users retain the option to disable the memory function.
Removing Barriers in Human-AI Interaction
Developers can use GigaChat Ultra to build applied AI products and services, including running code directly within the interface.
Anton Frolov, senior vice president and head of generative AI development at Sber, described the update as “a step toward a multi-agent AI assistant.”
The long-term goal is to create a future where neural networks replace traditional applications. GigaChat Ultra is one of the largest models globally and has been fully developed and trained in Russia.
The model was trained in three stages, beginning with academic texts, mathematics, and programming, followed by refinement of a specialized model and additional training on practical examples.
