Imagine a university where students have instant access to AI tutors, personalized feedback, and cutting-edge research tools. That's the promise of FLEXI, the FernUniversität in Hagen's experimental infrastructure designed to bring the power of open Large Language Models (LLMs) to higher education. One of the biggest hurdles facing universities today is the limited access to advanced LLMs. Commercial models are often expensive, raise privacy concerns, and offer little control over data usage. FLEXI tackles these challenges by implementing a locally hosted, open-source approach. This means universities can run powerful LLMs on their own servers, keeping data secure and avoiding costly subscription fees. FernUniversität's project aims to showcase the practical benefits and address the challenges of this approach. FLEXI utilizes readily available hardware and the open-source software, Ollama, to serve a variety of LLMs, allowing researchers to experiment with different model sizes and capabilities. Their setup demonstrates that even with moderate resources, universities can deploy potent LLMs. The team explored different models, prioritizing openness, language support, performance, and security. They also carefully considered benchmarks relevant to educational settings, such as German language proficiency and reasoning abilities, rather than just general knowledge. The results are promising. FLEXI has enabled faster response times, reduced costs compared to commercial cloud options, and significantly improved data privacy. The university currently monitors server load and energy consumption, providing valuable insights for future scaling. But FLEXI is more than just an infrastructure project; it's a glimpse into the future of learning. Imagine students interacting with AI-powered chatbots in Moodle, asking questions about lectures, and getting personalized tutoring. Researchers can access specialized models for their work, all within a secure university environment. While the project is still in its experimental phase, it has already identified key areas for improvement. Scaling to university-wide usage requires a robust operational plan, clear legal guidelines, and ethical considerations, including responsible AI usage policies. FLEXI demonstrates that the potential of LLMs in higher education is immense, but it also emphasizes the need for careful planning and responsible implementation. As open-source models continue to improve, projects like FLEXI pave the way for accessible, secure, and cost-effective AI tools, transforming how we learn and teach.