Skip to content

AI Models – Overview

This page provides an overview of the AI models available on the Evoya platform and explains how they are deployed and managed.

Fundamentals

AI models form the technical foundation for AI agents. The Evoya platform uses Large Language Models (LLMs) that can understand, process, and generate text, images, or other content.

Each AI agent uses exactly one language model. Which model is used is defined when configuring the AI agent. Additionally, default models can be defined at the organization level to serve as presets for newly created agents.

A key advantage of the platform is the ability to use multiple AI models in parallel and select them specifically based on the use case.

Model Types

The Evoya platform distinguishes between two fundamental types of AI models. This distinction describes the nature of the model and its licensing.

Open-Source Models

Open-source models are freely available language models whose source code is openly accessible. They can be operated flexibly and adapted to different requirements. Open-source models are particularly suitable for scenarios where transparency, control, or self-hosting are priorities.

Proprietary Models

Proprietary models come from external providers and are not openly accessible. They often offer specialized or particularly powerful features. A well-known example of a proprietary model is ChatGPT.

Processing Locations

AI models are grouped by their processing location. The location indicates where content is processed by the respective model.

Currently, models are available from the following regions:

  • Switzerland
  • Europe
  • United States

The processing location is a key decision criterion when selecting a model.

Default Models at Organization Level

Default models can be defined at the organization level to serve as presets for newly created AI agents:

  • Default language model
  • Default image model

These default models are automatically selected when a new AI agent is created.

When configuring an AI agent, the preset default model can be overridden and another activated AI model can be selected. This allows individual agents to be specifically adapted to different requirements.

The default embedding model is defined at the system level and cannot currently be customized.

Managing AI Models

In the Model Overview :material-open-in-new: AI models can be activated or deactivated. The models are grouped by their processing location.

  • Each model can be individually activated or deactivated
  • All models per location can be collectively turned on or off

Only activated models are available for selection when configuring AI agents.

Model Details

Detailed information can be displayed for each AI model. This includes:

  • Number of AI agents using this model
  • Context window size
  • Model parameters
  • Support for Tool Calling / Function Calling
  • Credit costs per 10,000 characters

This information helps in selecting the appropriate model for a specific use case.

Credits and Usage

The use of AI models is based on Credits. Credit consumption depends on the model used and the amount of text processed.

Credit costs are transparently displayed for each model.