KIT's KI-Toolbox: Your Gateway to Large Language Models

To gain access to the KI-Toolbox, a mandatory prerequisite is the successful completion of the self-learning course 'Understanding and Applying Generative AI.' This course is available on KOALA for KIT employees and on the ILIAS learning platform for students. Upon successful completion and a subsequent login to the KI-Toolbox, access will be automatically granted. Initial access is available to KIT employees, KIT students (from April 2026), and individuals with a guest and partner account. Before commencing usage, it is crucial to review and accept the KI-Toolbox's terms of use and privacy policy, which are presented during the first login. A quick start involves navigating to the KI-Toolbox URL, logging in with your KIT account, accepting the policies, creating a new chat, selecting a desired KI model from the dropdown menu, and entering your query. The system prompts users with 'How can I help you today?' for input.

The KI-Toolbox provides a transparent cost structure for its external AI models, with prices listed in US dollars per 1 million tokens. The cost is divided into input tokens (your query and context) and output tokens (the AI's response). As a general guideline, 100 tokens approximate 75 words. For instance, 'GPT-5 mini' from Azure costs $0.28 per 1 million input tokens and $2.20 per 1 million output tokens. 'GPT-5.4' and 'GPT-5.5' have higher costs, especially with context lengths exceeding 272,000 tokens, where prices for 'GPT-5.4' become $5.50 in and $24.75 out, and for 'GPT-5.5,' $11 in and $49.50 out. The current exchange rate of approximately $1 ≈ 0.85 EUR is used for conversion. Local models hosted at KIT do not incur token-based costs.

For developers and advanced users, the KI-Toolbox offers an OpenAI-compatible API for programmatic access to its AI models. A personal API key can be generated directly within the KI-Toolbox by navigating to 'Settings' (bottom left menu), then 'Accounts,' and selecting 'API-Key.' Users are strongly advised to treat their API key as a sensitive password and protect it from unauthorized access. If there is any suspicion of a compromised API key, it should be immediately regenerated through the KI-Toolbox. For external tools, the base URL for the API is typically 'https://ki-toolbox.scc.kit.edu/api.' The specific model IDs can be found in the model table or by hovering over the model selection in the KI-Toolbox to view tooltips. Detailed instructions for API usage are available in the ZML's guide.

The KI-Toolbox facilitates collaboration through its group functionality, allowing for the sharing of various artifacts such as models, knowledge bases, notes, and prompts among multiple users. While groups cannot be created directly within the KI-Toolbox interface, they can be synchronized via the SCC Group Management system. Groups are automatically created in the KI-Toolbox with names following the pattern '<OE>-ki-toolbox-<name>' when a user who is a member of such a group logs in. This synchronization ensures that group memberships are kept up-to-date, with new groups being created, new memberships added, and outdated group memberships removed automatically.

For any inquiries or support related to the KI-Toolbox, users can reach out to the dedicated contact email: ki-toolbox@scc.kit.edu. Additional resources are available to help users navigate and maximize their use of the platform. These include direct links to the KI-Toolbox (Open WebUI), guides and webinars provided by the ZML (Zentrum für Mediales und Wissensmanagement), and the self-learning courses 'Generative KI' on KOALA for employees and ILIAS for students. Users are also encouraged to consult the KI-Toolbox's usage terms and privacy policy for comprehensive information. The last modification date for the page was May 7, 2026.