Update on on-premise LLM service

Miek Gieben — Wed, 03 Jun 2026 14:55:08 +0200

In a previous news article we announced the availability of a local LLM with no connection to big tech. We haven’t really advertised it further, but we can say it has proven popular.

Part of the reason to launch was to discover bottlenecks, well, we’ve discovered bottlenecks.

Currently there are ~200 registrations and ~10 people online concurrently. Heavy users may have noticed that the service can be slow at times.

Plans

In order to make more efficient use of our server, we will:

New on-premise LLM service, available for all Science Students and Staff

Miek Gieben — Thu, 05 Feb 2026 09:12:56 +0100

At the start of this year we installed a high-performance server (512 GB memory, two high-end GPUs). This machine now hosts local large-language models LLMs, meaning the AI runs entirely on‑premise and your data never leaves our network.

What you can do today

Access the web UI at chat.science.ru.nl. Log in with your Science RU username (e.g., jdoe, not jdoe@science.ru.nl).
The service is powered by Ollama and Open WebUI, providing a clean, responsive chat interface.
Early tests show reliable speech‑to‑text (including Dutch) and the ability to review source code. A user experience that feels very close to the larger commercial models, even though we are currently using open‑source models a few generations older.

Current status - experimental

We don’t know how many concurrent users the system can support and how much persistent storage will be required.
No automated backups are in place; please treat the data you store there as temporary.

Uptime is not guaranteed. The service may need to be restarted or the server rebooted when we apply configuration changes or updates.

The setup will evolve as we gather feedback, so your input is valuable.

Which models are available?