PJSC Sberbank

The update allows corporate developers to integrate the AI tool faster

The solution will help businesses accelerate online analytics based on big data

Along with the model, Sber published tools that accelerate the training of diffusion models.

The AI analyst operates on the basis of the GigaChat neural network model.

Previously, the platform already offered mirrors for Java (Maven Central), JavaScript (NPM), and containers (Docker Hub).

O.K.E.A.N. was presented in May 2026 at the CIPR conference in Nizhny Novgorod.

It works and runs from the command line regardless of the development environment type and operating system.

AI agents will be able to conduct economic activities independently.

As part of the cooperation, the parties plan to implement analytical tools for operational, tactical, and strategic market analysis.

Sber plans to make AI more accessible and useful for businesses of all sizes.

"Marcus" is designed as an autonomous AI agent for Sber's marketing and communications department.

The system was developed by the subsidiary SberTech based on the open-source Russian ClickHouse DBMS and supports a SQL-like query language.

The platform allows delegating routine tasks to AI agents by describing their operational logic in the language of business regulations without developer involvement or changes to internal IT systems.