UNIT.City — місце, де люди працюють... КРАЩЕ! Обирай свій простір просто зараз 👉
Вікторія ГорбікAI Eng
10 April 2025, 13:35
2025-04-10
"We are not considering Chinese models and approaches." CTO of the AI Center of the Ministry of Digital Affairs on the development and probable release date of the national LLM
In early February, the Ministry of Digital Affairs launched the AI Center of Excellence, the first in Ukraine, which is also working on the development of a Ukrainian large language model (LLM). Dmytro Ovcharenko, CTO of the AI Center of Excellence and AICTO of the Ministry of Digital Affairs, reported that it is now known about the development of the model and when it can be expected to be released.
In early February, the Ministry of Digital Affairs launched the AI Center of Excellence, the first in Ukraine, which is also working on the development of a Ukrainian large language model (LLM). Dmytro Ovcharenko, CTO of the AI Center of Excellence and AICTO of the Ministry of Digital Affairs, reported that it is now known about the development of the model and when it can be expected to be released.
About the National LLM
The national large language model (LLM), according to Dmitry Ovcharenko, is usually based on open-source architectures such as LLaMA, Mistral, or Gemma, and is supplemented with specific national language corpora, the CTO of the AI Center of Excellence said in an interview for DOU.
About development
The team has not yet started the direct development of the model. According to AICTO, the concept development stage is currently underway: we are defining the tasks, forming the organizational structure, assembling the team, estimating the budget and timeline, looking for partners and mechanisms for involving scientists, universities, and business. In addition, he noted that the development process will definitely be as public as possible.
«The only thing I can say for sure is that we are not considering Chinese models and approaches,» he said, adding that this will be pre-training on an existing architecture, not developing from scratch. The developers will focus on small language models (1-5 billion parameters) and medium-sized ones (12-16 billion parameters), taking into account the experience of Gemma and the latest versions of LLaMA.
About the dataset
To train the national language model, according to Dmytro Ovcharenko, news, Wikipedia, and other information collected and provided by the community and universities that have been collecting open sources in the Ukrainian language for years will be used. «There is also the „Malyuk“ dataset. It is one of the largest — 113 gigabytes of cleaned text. In addition to it, there are NER-UK, UA-GEC, BrUK, and others,» he adds. All data will be checked by experts — historians, linguists, and cultural figures.
About users
In addition, Dmytro Ovcharenko added that the Ukrainian model will be freely available to the non-profit sector — the state, universities, schools, and scientists. As for business, we are still thinking about the conditions.
About money
The team is currently looking for investors to finance the development of the national LLM. Dmytro Ovcharenko does not name the specific amount of necessary expenses, but according to him, from the experience of other countries, the budget can range from $1.5 to $8 million.
About deadlines and exit
The CTO of the AI Center of Excellence says that the team plans development according to the Roadmap. «Under ideal conditions, the average model should be released in nine months. That is, in late November-December 2025,» he says. But Dmitry also added that it is planned to release not only one model, but also guardrail, embeddings, tokenizer, that is, a whole ecosystem of certain models.
About plans
The team aims to become one of the top three countries in the world in developing and implementing AI in the public sector by 2030.
Dmytro Ovcharenko is confident that AI will create demand for new specialties and change traditional approaches to work in various industries.
The Ministry of Digital Affairs has launched Ukraine’s first AI Center of Excellence, which will be a center for integrating AI solutions. What is known about the structure and what has already been done
The newly created AI Center of Excellence from the Ministry of Digital Affairs in Kyiv is recruiting an R&D team. Which specialists are they looking for?
