PARIS, France, Nov 17, 2025 – Dragon LLM (formerly Lingua Custodia) has introduced Dragon, an AI architecture developed and trained on EuroHPC supercomputers including Leonardo and JUPITER. Dragon LLM won the European Large AI Grand Challenge in June 2024. The Dragon AI architecture uses a hybrid design built for long-form text and complex reasoning. This reduces energy usage and inference costs while maintaining output quality comparable to other language models.
A European First for Business
Dragon draws on research from global laboratories over the past two years. This architecture represents a streamlined AI that trains faster, lowers operating costs and supports deployment across organizations. The project also shows that startups and SMEs can access the resources needed to build competitive AI models.
Olivier Debeugny, CEO and founder of Dragon LLM, explains his vision: “Our goal is simple – to build responsible and accessible AI for European companies that delivers real return on investment. Europe can innovate and create sovereign models, and Dragon is proof of that. It’s a powerful symbol: a small French Tech company developing AI models that rival the performance of major international players – without raising any external funding. Dragon shows that Europe can build and own its own AI technologies.”
What Dragon Demonstrates
- Reduced computational requirements:
Dragon provides similar performance to other models while using one-third of the computing power. - Standard server deployment:
The model can run on conventional servers without requiring large GPU resources. - Exceptional energy efficiency:
European data centers are seeing higher energy use, and Dragon offers an alternative. Its training process is designed to use less data and reduce resource demands before the model is deployed. - Outstanding operational performance:
On similar hardware, Dragon supports twice the users while matching results from Chinese and U.S. models.
Source: Dragon LLM
About Dragon LLM
Dragon LLM, founded in 2011 under the name Lingua Custodia, develops large-language model (LLM) solutions for enterprise use. The company is based in Paris, France. It focuses on designing, fine-tuning and deploying domain-specific LLMs for industries such as finance, insurance and professional services. Its tools support back-office, front-office and report-generation workflows. Dragon LLM develops AI models that can operate on infrastructures and align with business requirements. The company won the Large AI Grand Challenge organized by the European Commission. Dragon LLM serves clients that include global banking, asset-management and investment-research organizations.