14 July 2025 - Artificial Intelligence The University's AI Platform: Open-source solution YoKI in regular operation
After a successful test phase, YoKI has now been launched as a regular URZ service. With YoKI's launch, students, researchers and employees now have permanent access to an internal university AI platform, currently with four large language models. All open-source models available on YoKI are operated exclusively on university servers.
The YoKI language models can offer support for study-, research-, and work-related tasks – without compromising data sovereignty and data protection.
Students, researchers and employees can access the platform via browser from the university network or using the university VPN. No login is required. YoKI currently consists of four different large language models (LLMs), each covering different areas of application. Using the available LLMs, you are able to:
- quickly process, structure or summarize large amounts of text data
- generate new texts, outlines or headings
- create, revise or translate texts in a wide range of languages
- generate, improve and expand code in numerous programming languages such as Python, C++, etc.
You can find more detailed information about the YoKI features as well as the current models and settings in the linked YoKI documentation.
An overview of the open-source models and their various specializations:
Currently, you can choose from four large language models. The model best suited to the task can be selected based on the area of application:
- Llama von Meta: a standard, high-performance system for text generation and processing
- DeepSeek: an alternative high-performance language model for general use
- Aya from Cohere: a multilingual language model for communication in over 100 languages
- Qwen: LLM with a wide range of programming languages and a focus on coding
AI infrastructure made@University: Success through teamwork
From the outset, the creation of YoKI has been an interdisciplinary collaborative project under the direction of Prof. Dr. Vincent Heuveline. Team members from the Engineering Mathematics and Computing Lab (EMCL) and the University Computer Centre were involved in the design, implementation and further development of the university platform. The new service was created in response to numerous requests from users.
Curious about how YoKI was developed? Marcus Buchwald, a doctoral student who played a key role in the YoKI project, provides insights into the project's history as well as background information on AI and LLMs in general.
Artificial intelligence (AI) has become firmly established in academia and is now indispensable. In order to optimize the use of AI at our university, we believe operating our own platform is the best solution. YoKI offers this option with a special focus on data sovereignty. However, YoKI is more than just a platform—it is an ongoing process in which we will take new developments in the field of AI into account and continue to build on them.
Prof. Dr. Vincent Heuveline, Director of the URZ and CIO of Heidelberg University
YoKI is dedicated to data security, an open-source approach and local processing
In addition to the performance of the LLMs, maintaining data sovereignty is a top priority in the development and operation of the platform. The open-source models made available through YoKI offer transparency and adaptability for university use. This allows them to be used as needed in studies, teaching, research and administration. Free from commercial ties, they are operated on university servers in compliance with data protection regulations. This means that control and responsibility for data and technology remain entirely in the hands of the university. Other advantages at a glance:
- Transparency and trust: Researchers can review the model architecture and training data of open-source LLMs to ensure reproducibility and compliance with ethical standards.
-
Absolutely no content from the YoKI platform leaves our campus:
- The prompts are not saved after the chat is closed and are not used for training and further developing the models.
- Information remains confidential, as third parties have no access to current chats.
- The service can only be used within the university network or using the university VPN service.
- Different models have different focal points to enable application-specific use.
Further development and expansion of university AI: New features and tools in the pipeline
The YoKI service is continuously being developed with the aim of providing a practical selection of powerful large language models. Existing models will be updated and adapted as much as possible or replaced by better models when they become available. In the future, YoKI will be expanded with additional open-source assistants and other features such as web search (currently not functional) and chat storage. A user-friendly and significantly improved user interface is also in the works. YoKI will also be gradually expanded to include additional functions such as image processing, image generation and multimodal tools, e.g., for uploading and downloading documents. Customized chatbots, such as the one for IT Service, are already being tested. In the future, it should also be possible to develop and deploy chatbots for specific questions in other administrative areas as well as in institutions and departments.
Check out the latest news and stay up to date! Click here to go to our newsroom
