GWDG LLM Service – Generative AI for Science

Starting from February 22nd, 2024, GWDG offers its own ChatGPT service, which runs on in-house hardware. All users with an AcademicCloud account can access the website and chat with a generative AI to support their work. The service is currently based on Intel’s NeuralChat 7b v3.1 and runs on GWDG’s HPC infrastructure. Using an internally-developed system, the web service is able to dynamically adapt and scale to large request volumes in order to guarantee a fast response time for as many concurrent users as possible.

The system was designed with special emphasis on data protection and privacy. The user’s prompts and responses are only stored locally within a session in order to protect the data. Therefore, all prompts and responses are lost when the website is closed. Only the number of requests per user is recorded in order to monitor the demand for this new service. The service is being actively developed to offer more accurate and complex models, and to allow integration with other platforms in the future. If you have any questions or feedback, please contact