Kyivstar and the Ukrainian Ministry of Digital Transformation have chosen Google’s open-source Gemma model to underpin a sovereign national large language model designed to reflect Ukraine’s language, culture and data security needs.
VEON has announced that Kyivstar, in partnership with the WINWIN AI Center of Excellence under the Ukrainian Ministry of Digital Transformation, has selected Google’s Gemma open-source model as the foundation for the development of Ukraine’s national large language model (LLM). The project will leverage Google’s Vertex AI infrastructure to provide the computing power and scalable environment required for advanced model training and deployment.
The decision marks a significant milestone in Ukraine’s Digital Transformation journey. By building a sovereign national LLM, Ukraine is laying the groundwork for Artificial Intelligence systems that are deeply rooted in its own language, culture, history and institutional context. The initiative reflects a broader ambition to ensure that the next generation of AI-powered services truly understands Ukrainian society rather than relying solely on global, generalised models that may overlook local nuance or context.
Building a sovereign Ukrainian LLM
As a strategic partner in the initiative, Kyivstar will take operational leadership of the Ukrainian LLM’s development. The national language model is designed to capture the full breadth of Ukrainian dialects, terminology, historical references and cultural context. Crucially, it will also meet strict data sovereignty and security requirements, ensuring that sensitive national data remains securely stored and processed within Ukraine.
This local-first approach is especially critical for sectors such as government administration, healthcare, financial services and education, where data sensitivity and regulatory compliance are paramount. By developing an LLM tailored specifically to Ukraine’s needs, the project aims to provide institutions with tools that are both technologically advanced and aligned with national priorities.
Why Google Gemma?
The selection of Gemma followed an extensive evaluation process that assessed multiple technical, linguistic and economic criteria. According to Kyivstar, Gemma offered the optimal balance between performance, efficiency and adaptability.
“Choosing Gemma gives us the best balance between performance and resource use as well as high-quality training for the Ukrainian LLM,” said Misha Nestor, Director of Digital Product Development, Kyivstar. “The model supports more than 140 languages including Ukrainian, features a context window of up to 128,000 tokens, offers multimodal capabilities and has a flexible architecture that can be adapted to different tasks.”
Gemma’s open-source nature was another decisive factor. By anchoring the national LLM on an open model from a global technology leader, Ukraine gains transparency, flexibility and long-term control over how the system evolves.
Strengthening international technology ties Google was selected not only for technical reasons but also for the broader strategic context of the partnership. The decision reinforces the technological and economic ties between Ukraine and the United States – ties that were further strengthened by Kyivstar’s listing on the Nasdaq in August 2025.
By collaborating with Google and leveraging its cloud and AI ecosystem, Ukraine is positioning itself within a global network of advanced AI development while maintaining national ownership over the resulting technology.
Ethical, linguistic and security considerations For the Ministry of Digital Transformation and the WINWIN AI Center of Excellence, controllability and ethical alignment were core considerations in choosing the base model.
“We are building the Ukrainian LLM on a ready-made open-source model. The main task in development is to train it on our unique data further,” said Danylo Tsvok, Chief AI Officer at the Ministry of Digital Transformation and CEO of the WINWIN AI Center of Excellence.
“When choosing a model, we focused on how well it already handles Ukrainian-language texts and how controllable it is during additional training. This will help minimise linguistic and ethical risks in our LLM.”
By starting with a model that already demonstrates strong multilingual performance, the development team can focus on refining Ukrainian-specific capabilities such as grammar, idioms, legal terminology and historical references.
VEON’s vision for localised AI For VEON Group, the Ukrainian LLM represents a continuation of its broader strategy to close the AI language gap across its markets.
“Kyivstar and the Ukrainian Digital Ministry have taken a major step forward today,” said Kaan Terzioglu, CEO, VEON Group.
“With a sovereign Ukrainian LLM, Ukrainian consumers, businesses and government institutions will be empowered to integrate cutting-edge technologies using augmented intelligence that truly speaks Ukrainian and understands Ukraine.”
The Ukrainian initiative follows earlier VEON-supported projects including KazLLM in Kazakhstan and an Urdu-language LLM initiative in Pakistan.
Google Cloud’s role
From Google’s perspective, the project highlights Gemma’s growing role as a foundation for national and domain-specific language models.
Krzysztof Kaziów, Director of Customer Engineering for Central and Eastern Europe at Google Cloud, said: “We are honoured that the Ministry of Digital Transformation and Kyivstar have selected Gemma as the foundation for the Ukrainian national LLM. This choice underscores Gemma’s strategic value, offering an optimal balance between performance and resources alongside its strong multilingual support.”
By deploying the model on Vertex AI, the project benefits from scalable infrastructure, advanced tooling for training and evaluation and enterprise-grade security features.
From foundation to applications The development roadmap for the Ukrainian LLM begins with optimisation. Kyivstar will first refine Gemma specifically for Ukrainian language use, including improvements to the tokenizer and training on carefully curated Ukrainian datasets.
Once completed, the Ukrainian LLM is expected to serve as a foundational layer for a wide range of AI-powered services across both the public and private sectors.
By training the model on Ukrainian data and embedding local context from the outset, Kyivstar aims to deliver outputs that are more accurate, relevant and actionable than those produced by general-purpose global models.
Together, Kyivstar, the Ministry of Digital Transformation, VEON and Google are laying the foundation for a new chapter in Ukraine’s AI journey, one where advanced technology is deeply aligned with national identity, security and long-term digital resilience. This initiative signals long-term commitment to responsible innovation, economic resilience, digital sovereignty and trusted Artificial Intelligence adoption nationwide.
