Source: Own illustration
The demand for powerful AI models at universities continues to grow rapidly. Applications such as chatbots and AI-supported services in the KI:connect environment are leading to more and more institutions using generative AI in research, teaching, and administration. At the same time, demand for locally operated open-source models is increasing.
RWTH Aachen University is currently testing several open-source models via KI:connect through the WestAI service center. The KI:Inferenz.nrw initiative was launched in January 2026 to provide universities in North Rhine-Westphalia with comprehensive access to open-source AI models in the future. It aims to provide universities in North Rhine-Westphalia with an infrastructure for the use of language models that is independent of the universities themselves.
Goals & Participants
The goal of KI:Inferenz.nrw is to establish a university-independent AI infrastructure for North Rhine-Westphalia. This will enable universities to use AI models without having to rely on commercial providers. The focus is on European models.
Participants include partners from the Open Source-KI.nrw project at Ruhr University Bochum and the University of Cologne, as well as RWTH Aachen University together with the AI service center WestAI. A prerequisite for using the service is the use of KI:connect (via API or web interface).
First AI Models From April 2026
From April 2026, the first language models for universities in North Rhine-Westphalia are to be made available as part of KI:Inferenz.nrw. The offer is aimed at all universities in North Rhine-Westphalia and can be used by all institutions that already use KI:connect or are preparing to introduce KI:connect.
In the first phase, the Mistral Small and GPT OSS models, among others, will be available.
RWTH Aachen University is one of the universities that has been able to use the joint offering from WestAI and OSKI.nrw on a trial basis since the beginning of March. This has made it possible to establish location redundancy for the GPT OSS and Mistral Small models, as well as to make additional embedding models available.
Step-By-Step Development of the AI Infrastructure
To begin with, KI:Inferenz.nrw will initially be provided as a prototype on existing hardware at redundant locations in Aachen and Cologne. The service will be expanded gradually and will follow a best-effort approach in the initial phase. Due to limited resources, there will initially be usage limits and restrictions on the available models. However, these are designed in such a way that typical applications such as chatbots can already be supported.
Communication regarding the use of the models will take place via the established channels of the KI:connect project.
In the future, the offering is to be further expanded as part of the planned KI:Expertisezentrum.nrw project to further support the use of AI in universities in North Rhine-Westphalia.
Adaptation of the Standard Model in KI:connect
KI:connect will also be adapted as part of this development. The previous standard model 4o mini will be replaced by GPT OSS. In doing so, RWTH Aachen University is actively supporting the use of self-hosted models within the university infrastructure.
Central Financing for Universities in North Rhine-Westphalia
There will be no initial provisioning costs for universities in North Rhine-Westphalia. The AI infrastructure will initially be financed centrally by the Ministry of Culture and Science (MKW) to enable rapid introduction and use.
Responsible for the content of this article are Denise Dittrich and Arlinda Ujkani.
Leave a Reply