Imagine a smart digital assistant that actually understands the nuances of smaller European languages and doesn't make things up. Instead of relying on giant American AI, this creates a transparent, European-made brain for computers. It's like building a high-quality library of local knowledge that businesses can use without worrying about privacy or bias.
European companies rely on non-European AI that often lacks accuracy in local languages and fails to meet strict EU privacy and transparency standards.
A 7.8B parameter multilingual model and the EuroEval platform for standardized testing across 8 languages.
If you are a pharmaceutical provider dealing with complex medical leaflets in multiple languages — this project developed BijsluiterBot that provides accurate medical information to users.
If you are a car manufacturer dealing with the need for localized, context-aware voice assistants — this project developed LLM applications for automotive use that improve human-machine interaction.
If you are a publisher dealing with youth literacy and reading engagement — this project developed BookBot that supports youth reading through AI interaction.
Based on available project data, the models are intended to be part of an open ecosystem, but specific commercial pricing is not mentioned.
Yes, the project utilized exascale computing to train a 7.8B parameter model on 2.3T tokens, demonstrating the ability to handle massive data scales.
The project aims to develop an open and democratized LLM ecosystem, though specific license types are not detailed in the report.
The project focuses on trustworthiness, factual correctness, and TDM-compliant data pipelines to ensure legal and ethical alignment with European values.
The use of LoRA-based adapters allows for efficient tuning, making it easier to adapt the model for specific low-resource languages or niche business needs.
The consortium is research-heavy with 11 partners across 6 countries, featuring 4 universities and 5 research organizations. While the industry ratio is low at 9% (only 1 industry partner and 3 SMEs), the presence of these partners suggests a clear path for technology transfer to end-users.
Contact Linköpings Universitet for technical specifications on the 7.8B model.
Contact us to find the right partner for implementing TrustLLM in your local market.
