Mobile operator Kyivstar together with the WINWIN AI Competence Center under the Ministry of Digital Transformation of Ukraine have identified Google Gemma as the base model for the development of a national large language model (LLM).
It is known that Kyivstar will be responsible for the operational part of the project, while Google Vertex AI infrastructure will be used for computing power. The press release says that the model will be adapted to the Ukrainian language space — including dialects, context and terminology.
The statement also said that the partners chose Google Gemma after a comparative analysis of alternatives. The key criteria were the quality of work with Ukrainian-language texts and the possibility of safe additional training of the model. According to Daniel Tsyvok, head of artificial intelligence at Mincifra and CEO of the WINWIN AI competence center, this will reduce linguistic and ethical risks.
Once the development is completed, the LLM will be able to be used to analyze regulatory and legal documents, as well as to create specialized solutions in education, finance, medicine and other industries.
Kaan Terzioglu, CEO of VEON Group — the company that owns Kyivstar — believes that the project will strengthen the intellectual capabilities of Ukrainian companies and government agencies.
The press release notes that the initiative builds on VEON’s broader strategy to bridge the AI language gap in its markets. The company has already launched large language models in Kazakhstan and Pakistan.
Recall, the creation of a large language model in Ukraine began in June 2025. In November, the head of the Ministry of Digital Transformation of Ukraine, Mikhail Fedorov, announced the launch of a joint initiative with NVIDIA. The first joint project within the partnership will be the creation of Diia AI LLM.

