In the past three years, the adoption of artificial intelligence by governments has increased by 150%. This is highlighted in the study «National Large Language Models» conducted by the Ukrainian AI community.

The primary reason for this growth is that countries are developing their local large language models, which assist in creating quality AI solutions for the public sector, healthcare, defense, education, and agriculture.

«The advancement of artificial intelligence is a top priority for us. We are implementing AI in various projects, from education to defense tech. This allows us to be more efficient. Our goal this year is to accelerate the integration of AI into our projects. Recently, we launched the WINWIN AI Center of Excellence, which will focus on the deployment of a national LLM that will underpin the development of state digital products and facilitate the integration of AI into Ukrainian businesses,» said Mykhailo Fedorov, Deputy Prime Minister of Ukraine.

A large language model (LLM) is a type of artificial intelligence that mimics human brain functions, capable of analyzing and generating text. These models serve as the foundation for AI chatbots like ChatGPT, Gemini, and DeepSeek.

To produce text, models are initially trained on vast amounts of data. Unlike English-speaking chatbots, national models are developed using data in the local languages of their respective countries, such as French, Bulgarian, and others.

The research analyzed 18 macro-regions representing 85% of the world's population, identifying the key advantages of creating models in local languages.

The experiences of various countries show that large language models:

• simplify the handling of regulatory documents (Albania, Serbia)
• enhance communication between authorities and citizens (Netherlands)
• optimize public procurement and translation of legal materials (Albania)
• improve agricultural efficiency through tailored farmer consultations (Nigeria)
• enhance medical processes: structure medical records, predict disease risks, and assist doctors in formulating medical recommendations (Sweden, Kenya)
• aid in defense by analyzing information from open sources (OSINT) and providing responses to military inquiries tailored to specific communication styles (USA, China)
• can specifically inform citizens about government assistance (North Macedonia)

By developing national models, countries create AI assistants and specialized chatbots for various groups, including doctors, students, farmers, civil servants, and others.

The study was conducted by the Ukrainian AI community in collaboration with the technology company OpenBabylon, the lang-uk community, and the Ukrainian Catholic University.