The University of Amsterdam has released a suite of updates to UvA AI Chat, its in‑house generative AI platform for students and staff. The improvements include expanded AI‑literacy learning modules, adoption of a more powerful default language model, and the addition of open‑weight models running on Dutch servers. These updates aim to boost performance, broaden accessibility and strengthen autonomy from commercial AI providers.
Since its launch in the 2025–26 academic year, UvA AI Chat has been continuously refined based on feedback from students and university employees. The updated e‑learning courses on AI literacy for both lecturers and students now include fresh examples, useful tips and responsible‑use guidelines tailored to the UvA community. A renewed introductory course guides users through all functionalities of the AI tool step by step.
Performance improvements are significant thanks to the adoption of gpt5.1 as the default language model in UvA AI Chat. This model, combined with optimized backend code and system prompts, offers faster response times and improved reasoning capabilities for both simple and complex tasks.
UvA AI Chat also now supports an open‑weight reasoning model (gpt‑oss‑120b) that runs on local servers in Watergraafsmeer in partnership with SURF. By hosting this model within the university’s own trusted environment, students and staff gain access to a broader range of AI models while reducing dependency on external tech platforms.
Another key enhancement concerns document handling. Earlier versions of UvA AI Chat did not always fully analyse uploaded documents. The updated system offers faster and more reliable processing, enabling users to upload material without errors and to have the content understood more accurately. Detailed instructions on effective use of document handling are available in the UvA AI Chat manual.
The updates reflect the UvA’s broader commitment to responsible AI use in education, including careful attention to privacy, data governance and integration with academic workflows. Contributors to the project, such as student‑turned‑tech lead Danny van den Berg, continue to refine the tool and highlight its key benefits for daily academic and administrative work.