We use LLMs to generate answers. We are large language model agnostic. Therefore we can add all existing large language models that are reachable via a public API. You can also add self-hosted models.
As standard, we provide you with Google’s Gemini models, which offer high performance at an affordable price. We have a processing contract with Google to be GDPR compliant (data will not be used to train new models, etc.). We make sure all models are hosted in Europe, preferably Germany as well. For an increased resilience we recommend using multiple models from different providers.Costs for LLM depend on the used model. The used models can easily be changed via our model controller. Soon you will be also able to change and manage the used LLMs in the admin panel. We generally need:
- API key
- Authentication key
- Endpoint
Model Availability and Deprecations
Providers retire models over time. We add new models as they become available and inform you before an existing model is switched off.Response Modes
In addition to the model itself, users select a response mode - Fast, Balanced or Precise - directly in the input field. Fast is the default and the right choice for most tasks. More precise modes spend more effort on an answer and therefore take longer.Response modes explained for end users
Click here to find out more about this topic.
Azure Available?
- If you want to use the ChatGPT-models, a processing contract with OpenAI is not necessary, as long as you have one with Azure
- Azure Portal -> search for Azure OpenAI
- In your Azure OpenAI interface: Manage Keys -> Read endpoint & keys and make them available to us
- You can make your network settings under Network. Even if the endpoint is public, it is secured by the authentication key. However, it is always advisable to restrict access. Please note that with Notes
- Create Azure OpenAI, API Key costs are charged in Microsoft Azure. Billing per token. Please provide us with the required information afterwards.