Ukraine is developing its own large language model using Google’s Gemma open-weight framework—a strategic move to establish AI sovereignty amid ongoing conflict and reduce reliance on costly foreign services.
The project is a collaboration between Ukraine’s Ministry of Digital Transformation and mobile operator Kyivstar, which will lead development. Initial training will use Google’s Vertex AI infrastructure before transitioning entirely to local data centers for national control.
Why Gemma#
Mykhailo Nestor, Kyivstar’s Director of Digital Product Development, explained the choice: “Choosing Gemma for the Ukrainian LLM gives us the best balance between performance and resource use, as well as high-quality training.”
The model supports over 140 languages including Ukrainian, features a 128,000-token context window, and offers multimodal capabilities. Danylo Tsvok, Chief AI Officer at the Ministry of Digital Transformation, said the selection focused on how well Gemma already handles Ukrainian-language texts and its controllability during training to “minimize linguistic and ethical risks.”
Building for Local Context#
The initiative aims to replace expensive foreign AI services like ChatGPT while avoiding dependence on Chinese models such as DeepSeek and Qwen—particularly important as Ukraine integrates AI into military operations including battlefield coordination and surveillance.
Training data is being gathered from over 90 sources, including government archives and war records. The model will support Ukrainian, Russian, Crimean Tatar, and minority dialects. The project will also improve the Ukrainian tokenizer and create benchmark tests for future applications.
Security is paramount. Kyivstar has implemented backup infrastructure to resist Russian cyberattacks, with features being developed to protect against prompt injection and other AI-specific threats.
Part of a Larger Strategy#
This follows VEON’s broader push to close the AI language gap across its markets—after launching KazLLM in Kazakhstan in December 2024 and an Urdu LLM initiative in Pakistan. Kaan Terzioglu, CEO of VEON Group, stated: “We have a responsibility to bring the benefits of augmented intelligence to the countries we serve—through large language models trained not only on words, but on local context.”
Kyivstar, which listed on Nasdaq in August 2025, plans to invest $1 billion in Ukraine between 2023-2027. The Ukrainian LLM will serve as the foundation for AI-powered services across government, healthcare, education, and financial services.