Исследователи Nazarbayev University представили уникальную языковую модель KAZ-LLM

Researchers from the Institute of Intelligent Systems and Artificial Intelligence (ISSAI) of Nazarbayev University (NU) presented to the Head of State Kassym-Jomart Tokayev a large language model of the Kazakh language ISSAI KAZ-LLM, developed on the basis of a neural network. This is the basis of the Kazakh GPT chat. Application of the unique KAZ-LLM language model The language model can be applied everywhere: from translations of the Kazakh language and content generation to processing of large texts. Adapted to the country's unique multilingual and multicultural context, the model is designed for Kazakh, Russian and English with additional support for Turkish, bridging language gaps and advancing generative artificial intelligence for resource-constrained languages. The institute's team collected, processed, synthesized and translated more than 150 billion tokens (that is, words or subwords), providing reliable language performance. The level of training of this model is very high and competes in results in Kazakh, Russian and English languages ​​at the level of world leaders in the field of artificial intelligence. In addition, the language neural network provided practical experience to local IT talent

there, strengthening the nation's artificial intelligence capabilities. The project contributes not only to the creation of an advanced artificial intelligence tool, but also to the growth of the number of Kazakhstani specialists in the field of artificial intelligence. Kazakh researchers participated in the development. Collaboration with leading Kazakh institutions has resulted in the creation of benchmarking tools and datasets tailored to the Kazakh language with the help of linguists and advanced machine translation techniques. Kazakhstan's quest for innovation The project, funded by the NU and NIS endowment fund, began in April this year. Data were collected from publicly available sources, including Kazakh websites, news articles and online libraries. Data provided by various organizations was also used. "This model reflects Kazakhstan's desire for innovation, self-reliance and growth of its technology ecosystem. Our team has prepared two versions of ISSAI KAZ-LLM with 8 billion and 70 billion parameters, built on the Meta Llama architecture and optimized for high-performance systems and resource-constrained environments. Models are released under the CC-BY-NC license,

which are available for non-commercial use on the Hugging Face website, promoting global academic and research collaboration. This way, developers will be able to download and run our model both on complex servers and on laptops,” said ISSAI Director, NU Professor Hussein Atakan Varol. Photo: Nazarbayev University press service