Tilde has significantly improved its open-source large language model TildeOpen , which can now simultaneously process eight times longer texts than before, thereby greatly expanding its capabilities in handling large volumes of information. The improvement was developed using the European supercomputer LUMI.
“This is a significant improvement to the existing TildeOpen LLM, making it an even more powerful base model for further applications. We remind you that this is not a ready-made chatbot or an end-user solution, but a flexible foundation that can be customised, expanded, and integrated into various artificial intelligence solutions in businesses and the public sector,” says Artūrs Vasiļevskis, Chairman of the Board of Tilde.
He points out that the increased context size allows the model to process significantly more information at once, improving its ability to handle long and complex texts where information may be scattered in various places.
This expands the model’s usability as a base for various AI solutions that require a broader understanding of context.
Investment in European artificial intelligence continues
TildeOpen is a secure and reliable artificial intelligence base model trained on data from 34 European languages. The TildeOpen base model has been available since September 2025, and since then, the Tilde research team has continued its development to make the model increasingly capable and suitable for the real needs of organisations.
“Europe needs its own reliable artificial intelligence that understands European languages, allows secure data management, and is also suitable for sectors with stricter security requirements. TildeOpen is our contribution to the development of such artificial intelligence in Europe, and we continue to improve it so that European organisations have a stable and reliable foundation for their artificial intelligence solutions,” says A. Vasiļevskis.
Most popular artificial intelligence solutions are designed for the English language, so they often work less accurately in other languages. TildeOpen is specially tailored to European languages, including the Baltic region, providing higher quality language usage.
It is important that the model can be deployed within the organisation’s internal infrastructure, thereby maintaining full control over the data and ensuring compliance with European data security requirements. This is especially important at a time when the demand for artificial intelligence solutions that ensure data sovereignty and also support less represented languages is increasing in Europe.
The model, with over 30 billion parameters, was developed following Tilde’s victory in the European Commission’s Large AI Grand Challenge.
Curious about implementing AI in your organisation?
Contact us today and see how our solutions can improve your workflow
