Sber Is Teaching GigaChat to Speak Tatar – and 20 Other Indigenous Languages

A Scientific Partnership With Cultural Stakes

Alexander Vedyakhin, First Deputy Chairman of the Executive Board of Sberbank, announced the launch of the project during a visit to Kazan. Sberbank and the Academy of Sciences of the Republic of Tatarstan signed an agreement aimed at expanding the linguistic capabilities of GigaChat.

Researchers from the Academy will provide linguistic expertise and archival materials. They will also review training datasets and help the neural network master the specifics of Tatar grammar. The goal is for GigaChat to generate original text in Tatar rather than simply translate content from Russian.

Building a Literary-Grade Training Corpus

To train GigaChat, developers assembled a large-scale corpus of Tatar-language texts. It includes classical literature, contemporary press, textbooks and academic publications. This approach exposes the model to the full range of linguistic expression – from everyday speech to poetry, from humor to formal administrative language.

The corpus is continuously updated. Specialists from Sber and the Academy curate materials that reflect living, contemporary language. Without sufficient high-quality data, a neural network cannot produce coherent, contextually relevant responses. The project addresses what developers describe as a digital data gap affecting minority languages.

Together with leading scientific institutions in the country, we are working to preserve the languages of Russia’s peoples for future generations. Artificial intelligence technologies are a tool that allows us to restore what has been lost and safeguard cultural heritage, which is essential for maintaining national sovereignty and identity. Residents of Tatarstan, one of the largest ethnic communities in the country, actively use GigaChat in Russian. That is why we chose Tatar as one of the first national languages for training the model

Alexander Vedyakhin

First Deputy Chairman of the Executive Board of Sberbank

Expanding AI to Reflect Russia’s Linguistic Diversity

Tatar is the second most widely spoken language in Russia after Russian. More than 5.3 million people identified it as their native language in the most recent census. Yet many digital services offer limited or no support for it. GigaChat is positioned to become one of the first AI assistants capable of holding full conversations in Tatar.

Residents of Tatarstan will be able to interact with the chatbot in their native language. This is particularly significant for older generations and rural communities, where Tatar remains the primary language of communication. In this context, AI functions as a bridge between traditional culture and modern digital infrastructure.

From Tatar to Twenty Languages

GigaChat is currently being adapted to support twenty languages spoken across Russia. In addition to Tatar, these include Udmurt, Altai, Bashkir, Buryat, Veps, Hill Mari, Ingush and Komi, among others. The Russian online encyclopedia RUVIKI has contributed more than 1.4 million texts in these languages to support model training.

The initiative spans languages from the Caucasus to the Russian Far East. It represents one of the most ambitious efforts to digitize the country’s linguistic heritage. Many of these languages face the risk of decline, and AI could become an infrastructure layer for their long-term preservation.

AI Sovereignty and Global Ambitions

The project signals that Russian developers are building and adapting neural network tools to serve the country’s own linguistic ecosystem.

If successful, Russia could become one of the few countries to deploy a domestically developed neural network supporting dozens of national languages. Countries with multilingual populations, including India and nations across Africa and Latin America, may find such a solution relevant to their own digital inclusion strategies.