AI-SANA: KazNU will complete the project to create a large language model for the kazakh language this year

Scientists at Al-Farabi Kazakh National University are completing a project to create large language models aimed at the technological development of the Kazakh language. The scientific project, which began in 2024, will be fully completed this year.

As part of the project, a significant amount of textual data in Kazakh was collected, cleaned, and structured. Based on this data, several large language models were developed, trained with Kazakh as the primary focus. The current versions of the models have 0.6 billion, 1.7 billion, and 4 billion parameters, respectively. One of the important technical features of the developed models is the ability to call external tools, or tool calling. This enables the models to interact with digital tools and perform complex tasks step by step. This function is of particular importance for the further development of multi‑agent and autonomous systems.

According to the project leader, Madina Mansurova, dean of the Faculty of Mechanics and Mathematics at Al‑Farabi KazNU, the project’s goal is not limited to training language models. “Along with developing large language models with priority support for the Kazakh language, we are creating multi‑agent systems that operate on their basis. This will make it possible to expand the qualitative use of the Kazakh language in artificial intelligence technologies, improve data security, and strengthen our country’s technological independence”,- Madina Mansurova noted.

To implement the project, a scientific consortium was established, which included Al-Farabi Kazakh National University, Nazarbayev University, the Institute of Information and Computing Technologies, the A. Baitursynuly Institute of Linguistics, and the National Scientific and Practical Center “Til‑Qazyna”. To put the developed technologies into practical use, the researchers launched three platforms: “AI MamaCare”, “Farabi Flow” and “AI KazMedia”. All three solutions are hosted on local servers. The information is not shared with external services, which ensures data security and allows the systems to be used within the internal infrastructure of organizations.

“AI MamaCare” is designed to support mothers and medical professionals during the postpartum period. The platform helps to conduct daily monitoring of women’s health status, track changes in health indicators, and provide the necessa
ry information to doctors.

“Farabi Flow” is an intelligent platform for the university’s digital services. With its help, applicants, students, and staff can quickly obtain the necessary information. The platform also supports the university’s scientific activities by allowing data on researchers, scientific publications, and key indicators to be systematized and analyzed in a single environment.

“AI KazMedia” is designed for journalists and representatives of the media. The platform helps optimize editorial processes and quickly create informational content in various formats. Currently, all three platforms have successfully passed testing and been put into practical use.

In the future, scientists at KazNU plan to continue developing new digital solutions and applications based on artificial intelligence for various industries.

Press-service of Al-Farabi Kazakh National University