Beta testing of national LLM planned for spring 2026 – 1st Dpty PM
Beta testing of national LLM (large language model) is planned to be launched in spring 2026, First Deputy Prime Minister for Digital Transformation Mykhailo Fedorov said.
"And the name for the Ukrainian LLM will be chosen by Ukrainians themselves — in January we will start voting in Diia," he wrote on the Telegram channel on Wednesday.
According to him, the first text database for LLM training will be formed in January. The tokenizer, which divides words into elements, will be improved to enable fast and productive language processing. It will also have its own benchmarks for assessing quality.
Fedorov explained that data collection for LLM training is currently underway. "High quality requires more than information from the internet, so we work with state bodies, media, universities, and other institutions. These sources provide terabytes of unique data," noted the first vice prime minister.
He added that specialists are working on creating language model tests to help evaluate and improve quality.
"In parallel, we are creating a legal framework to responsibly work with the data provided to us for training the model," Fedorov emphasized.
The technical partner of the project is the largest Ukrainian mobile operator, Kyivstar, which is financing the development of a national large language model and its subsequent transfer to the state.
The First Deputy Prime Minister specified that a group of independent experts from various fields will test the quality of the development. Specifically, the model will be evaluated based on technical quality, ethical considerations, user safety, Ukrainian language proficiency, and understanding of the national context.
