Sberbank, Russia’s largest financial institution, is preparing to launch a new iteration of its GigaChat large language model (LLM), equipped with enhanced reasoning capabilities and the ambition to power scientific research and complex problem-solving across science, math, and coding.
This strategic move, revealed by First Deputy CEO Alexander Vedyakhin in an interview with Reuters, underscores Sberbank’s push to close the AI innovation gap between Russia and global leaders in the United States and China.
The upcoming GigaChat release will feature reasoning functions, setting the stage for high-level scientific inquiry, Vedyakhin said, adding that he is already testing the beta version of the advanced model.
These new reasoning-enabled LLMs are designed to tackle difficult tasks previously beyond the reach of traditional models, keeping pace with offerings from industry heavyweight OpenAI, which integrated similar capabilities into its flagship product last September.
Vedyakhin noted that, despite a lingering technological lag of six to nine months behind top-tier AI competitors, GigaChat is gaining traction in Russia, owing to superior language localization, robust performance, and hosting on domestic cloud servers that appeal to local corporate clients concerned about data sovereignty.
Presently, some 15,000 Russian companies are utilizing Sberbank’s GigaChat, positioning it as a front-runner in the country’s fast-evolving AI ecosystem.
