AI platforms at UU

Within a single environment, you have access to multiple language models: commercial models via Azure OpenAI and Mistral, and open-weight models running on Dutch SURF servers.

You choose which model best suits your task. The more complex the query, the more computing power you can request. A complex query often requires a better model. But that also consumes more energy. Often, a simple model is sufficient, allowing you to save energy. Energy consumption is indicated by the number of lightning bolts.

What can you do with it?

• Chat with multiple language models and choose for yourself which model suits your question
• Add your own documents and files to a conversation and ask questions about them (PDF, Word, Excel, PowerPoint, CSV, text files and dozens of other formats)
• Use and create your own personas: assistants with their own instructions and context for a recurring task, a subject or a work process
• Work on projects, with fixed knowledge and agreements that the model carries over into every conversation
• Build your own prompt library, alongside a set of standard prompts for common tasks
• Search the internet for up-to-date information
• Generating images and analysing existing images
• Collaborating on text, code or visualisations in a separate workspace, with version history and a download option
• Using study mode: the model explains step by step and asks questions, rather than giving the answer straight away
• Using voice: speaking instead of typing, and having answers read aloud
• Have diagrams created, such as flowcharts and timelines
• Analyse data, for example a CSV file, and run code within the environment
• Generate interactive practice questions and quizzes based on your own material
• Review conversations; your conversation history is saved within the UU environment

How do you gain access?

This service will be available to students and staff from the start of September.

What should you bear in mind?

The output from a language model may be incorrect or incomplete. You remain responsible for checking and using the results. Use the platform within the university’s frameworks and guidelines for responsible AI use. A basic level of AI literacy is a prerequisite; Utrecht University offers training courses for this.

How long will it be available?

UU AI-Chat will be available from the 2026–2027 academic year and will be provided and managed by Utrecht University as a regular service.

Data and privacy

The service runs within Utrecht University’s Azure environment and data is processed within the EU. Your data remains within the university’s own environment and is not used to train AI models. Your chat history is stored within the UU environment.

In addition to commercial models, EduGenAI also provides access to open-source models hosted by SURF itself on Dutch infrastructure (the SURF AI Hub). This offers an option that is not dependent on commercial providers.

The more complex the query, the more computing power you can request. A complex query often requires a more advanced model. But that also consumes more energy. Often, a simple model is sufficient, which saves energy. Energy consumption is indicated by the number of lightning bolts.

What can you do with it?

• Chat with different language models and choose for yourself which model suits your query, including based on speed and energy consumption
• Add your own files to a conversation and ask questions about them (PDF, Word, Excel, PowerPoint and CSV)
• Create and use personas: AI assistants with their own role, expertise and instructions
• Work on projects, where you can group conversations, prompts and personas together by task
• Share personas and prompts with others via groups
• Build a prompt library with prompts you use frequently
• Use study mode: the model asks questions and provides feedback, rather than giving the answer straight away
• Collaborate on text or code in a separate workspace (artifacts)
• Have diagrams generated, such as flowcharts and Gantt charts
• Analyse CSV files, generate charts and run Python code
• Generate images
• Search the internet for up-to-date information
• Use voice: speak instead of typing
• Customise the platform via the settings: interface language, theme, default model, conversational style, custom instructions and memory

How do you gain access?

You can request access using the forms on this page.

What should you bear in mind?

EduGenAI is a pilot project and still under active development; not all features are fully functional yet. Please do not share any privacy-sensitive information in EduGenAI at this stage. For example, the web search function is still being developed and the Python functionality may contain errors. Very large documents (more than around a hundred pages) may be too complex to process properly.

How long will it be available?

EduGenAI will be available as a pilot project throughout the 2026–2027 academic year.

Paid version:

• Full access to all Vibe features, including advanced coding (Vibe for Code), agentic workflows, and integrations with development environments (VS Code, etc.).
• Higher message limits and the ability to perform complex tasks and in-depth reasoning (compared to the free version of Mistral).
• Shared workspaces with admin controls and team management.
• Integration with existing systems (e.g. Slack, custom connectors via the Model Context Protocol). These types of integrations are being explored during the three-year pilot period.

For developers

Via the Studio API, you can build your own AI applications: agents and workflows, search within your own documents (RAG), audio and embeddings. You can program via the CLI, from VS Code or via the web interface. You can build your own connections to other systems via the Model Context Protocol.

How do I gain access?

You’ll soon be able to read more here about how to apply for access.

How long will it be available?

The pilot will run for three years.

Data and privacy

Mistral is a European provider. In the paid version, all data is processed within the EU.