Soofi: European sovereign LLM trained in 2 months (huggingface.co)

A team of European researchers has successfully developed "Soofi," a sovereign large language model (LLM) that was trained in just two months. This achievement marks a significant advancement in the realm of AI/ML, demonstrating that sophisticated models can be developed rapidly while retaining a focus on local data and regulatory compliance. Soofi's training leveraged a diverse dataset that emphasizes European languages and cultural contexts, positioning it as a tailored solution to address specific regional needs. The significance of Soofi lies not only in its rapid development but also in its potential to enhance data sovereignty and security in AI applications across Europe. By prioritizing local data, the model aims to mitigate privacy concerns associated with larger, predominantly U.S.-trained models. Technically, Soofi employs state-of-the-art techniques in unsupervised learning and transfer learning to optimize performance despite the limited training timeframe. The implications of this development could reshape the landscape of AI governance and promote the adoption of indigenous models, thus empowering European organizations to retain control over their data and AI systems.