Ukrainian Archive Service Provided 10 Terabytes of Data for the Language Model “Syaivo”
The State Archive Service of Ukraine transferred 10 terabytes of data for the development of the national language model “Syaivo”. This is the first such case, marking the beginning of a new era in the development of Ukrainian digital services.
The transfer includes historical sources, manuscripts, state documents, laws, court decisions, media materials, and dictionaries, equivalent to 70 thousand books.
The initiative aims to eliminate issues related to English translation responses provided by global AI assistants. More than 50 partners, including media, universities, and libraries, are involved in creating the language model. The full list is planned to be published after the model’s launch.
The foundation for LLM training was the Gemma 3 model from Google, adapted for the Ukrainian language and national context. Valeria Koval, Deputy Minister of Digital Transformation, emphasizes that “Syaivo” will facilitate the automation of state services, improve their quality, and assist in decision-making defense efforts. The open beta testing is scheduled for late spring, with initial access for governmental institutions and researchers.
| Data Volume | 10 terabytes |
| Equivalent in Books | 70 thousand |
| Partners | More than 50 |
| LLM Foundation | Gemma 3 from Google |
| Beta Testing Date | Late Spring |
