Understand and Select Language Models (LLMs)

Behind Ayunis Core are so-called language models (also called LLMs – Large Language Models). These models generate the texts and responses you receive in chat. Here you’ll learn what you need to know about them.

What is a Language Model?

A language model is software trained on vast amounts of text. Through this training, it has learned how language is structured and can independently generate texts, answer questions, summarize content, or translate. The responses are generated word by word based on probabilities – the model “calculates” which word fits best next.

Important: AI is Not a Reference Book

Language models generate texts based on probabilities – not by looking things up in a database. Always check AI responses for accuracy before using them. Specifically, this means:

• Always verify facts – The AI can provide plausible-sounding but incorrect information
• No real-time knowledge – Without internet research enabled, the AI only knows its training data
• Same question, different answer – The AI can answer the same question differently

Available Models

Which language models are available to you is determined by your administrator. The default model for your organization is also set by the administrator. The following models are currently available:

Model	Provider	Strengths
Claude Sonnet 4.5	Anthropic	Excellent text quality, careful and nuanced responses, strong with longer texts and analyses
GPT 5.2	OpenAI	Versatile, fast across a wide range of tasks
Mistral Medium 3.1	Mistral	European provider, strong performance with German-language texts, fast responses

Which Model Should I Choose?

When creating a chat, you can select a language model from the models enabled by your administrator.

Your Organization’s Default Model

Your administrator has set a default model for your organization. This model is automatically suggested when you start a new chat. It is suitable for most everyday tasks. You can set your own default model in your chat settings, which will override the organization setting for you.

Guidance

Task	Recommended Model
General questions, short texts, quick answers	Your organization’s default model
Complex analyses, detailed reports, nuanced evaluations	Claude Sonnet 4.5
Versatile tasks	GPT 5.2
German-language texts, quick drafts	Mistral Medium 3.1
Sensitive data	Any model with anonymous mode enabled

Tip

If you’re unsure, simply use your organization’s default model. You can switch the model at any time during a chat if you’re not satisfied with the result.

Switching Models During a Chat

You can also switch the language model within an active chat:

1. Click the model selection in the input area
2. Select another model from the list
3. Your next message will be answered with the new model

The previous conversation history remains intact.

Next Steps

• Prompting: Basics & Best Practices – How to get the best out of the language model