Ukrainian Platform for Domestic LLMs Runs on De Novo Cloud
2026-05-15
Lapathoniia, an infrastructure platform for hosting and integrating Ukrainian-developed large language models (LLMs), has been launched in Ukraine. The service runs on the De Novo cloud.
Ukraine has taken another significant step toward digital sovereignty. On May 15, 2026, Lapathoniia, a Ukrainian platform for domestic AI models, was launched. The service provides API access to Ukrainian LLMs. Currently available models include MamayLM and Lapa LLM, with developers planning to expand the lineup with other locally developed solutions in the future.
The platform is built on De Novo’s secure cloud and runs on NVIDIA H100 GPUs, which are widely regarded as the gold standard for demanding AI workloads. Academic support for the project is provided by the Department of Artificial Intelligence Systems at Lviv Polytechnic National University.
“Now any developer building an AI system for the Ukrainian market knows there is a service capable of hosting their solution. This should encourage more such developments. Companies and government institutions that need data to remain within Ukraine finally have an easy-to-integrate option for using LLMs,” Bohdan Didenko, one of the project’s co-founders, wrote on LinkedIn.
According to the developers, Lapathoniia provides everything needed to integrate Ukrainian LLMs into projects: an OpenAI-compatible API, models optimized for Ukrainian-language queries, contexts, and use cases, API keys, usage limits and quotas for teams and testing, support for streaming generation in interactive applications, and more. Currently, two LLMs are available: Lapa, a 12B Instruct model with vision support, and Mamay, a 12B Gemma 3 Instruct model optimized for the Ukrainian language. Both support context windows of up to 32,000 tokens.
Another key advantage is that all infrastructure and data are hosted in Ukraine, within De Novo’s secure and certified data center.
“Ukrainian LLMs need local infrastructure that can handle complex inference workloads and ensure data control within the country. For Lapathoniia, the De Novo cloud based on NVIDIA H100 serves as that foundation — delivering the performance required for large contexts, multimodal scenarios, and API-based integration,” said Dmytro Fedorenko, Director of AI at De Novo.
Lapathoniia is currently in closed beta, and early users have already received API keys to test the service. The team plans to make the platform publicly available in the future.