Managing Models
Available models can be configured per environment. These models will only work if the related API Key is also entered (and functioning) for the environment.
Available Models
Currently, the following models can be configured:
| Category | Models |
|---|---|
| Language Models | Claude Sonnet 4.5, GPT 5.2, Gemini 3 Pro, GPT 4.1 Mini, Claude Haiku 3.5, Gemini 3 Flash |
| Image Models | GPT-Image-1, DALL-E 3, SDXL, SD3.5 |
| Video Models | SVD (Stability Video Diffusion) |
AI-Public aims to implement newly available models as quickly as possible within the app. We will keep administrators informed as soon as new models can be implemented.
Model Settings
Models have the following settings:
| Setting | Description |
|---|---|
| Available to users | Yes/no - determines whether users can use this model |
| Available for Assistant API | Yes/no - only for OpenAI language models |
| Available for document chat | Yes/no - recommended for models with large context window |
| This is a thinking model | Yes/no - enables thinking mode functionality |
| Fast model (alternative) | Dropdown - selects the linked fast model |
Models can be ordered by dragging a row in the sorting column. This order will be maintained when loading models.
Thinking Model Configuration
A powerful feature is the thinking model. This allows users to switch between a fast model for daily tasks and a powerful model for complex questions.
Setup
- Open the edit dialog of a powerful model (e.g., Claude Sonnet 4.5)
- Check "This is a thinking model"
- Select a "Fast model" from the dropdown (e.g., Claude Haiku 3.5)
- Save the model
How Does It Work?
When a user selects this model:
- By default, they communicate with the fast model
- By clicking the 💡 icon, they switch to the thinking model
- The user can switch between both at any time
Benefits
| Benefit | Description |
|---|---|
| 💰 Cost-efficient | Fast models consume fewer tokens |
| 🚀 Faster | Quick answers to simple questions |
| 🧠 Flexible | Switch to thinking model for deeper analysis |
| ✨ Best of both | Combine speed with power |
Recommended Combinations
| Thinking Model | Fast Model |
|---|---|
| Claude Sonnet 4.5 | Claude Haiku 3.5 |
| GPT 5.2 | GPT 4.1 Mini |
| Gemini 3 Pro | Gemini 3 Flash |
Azure OpenAI Models
In addition to standard OpenAI models, it is also possible to use Azure OpenAI models. This is ideal for governments and public institutions that already have an Azure subscription or need to meet specific government compliance requirements.
Requirements
To use Azure OpenAI, you must first enter the following details in the API Keys of your environment:
| Setting | Description |
|---|---|
| Azure OpenAI API Key | The API key from your Azure OpenAI resource |
| Azure OpenAI Instance Name | The name of your Azure OpenAI instance (e.g., municipality-openai) |
Adding a Model
When creating a new model:
- Click "New model"
- Enter
Azurein the Company field - An additional field appears automatically: Azure Deployment Name
- Enter the name of your Azure deployment (e.g.,
gpt-4oorgpt-35-turbo) - The Identifier can be the same as the deployment name
- Data residency: Government data stays within the EU or Azure region of your choice
- Government security: Integration with Azure Government and network security
- Existing contracts: Use your existing government contracts with Microsoft
- Compliance: Meets government compliance standards (FedRAMP, GDPR, local regulations)
Help with Setup
AI-Public will provide assistance in setting up the models. Sometimes model identifiers need to be adjusted because a new version of the model is available. We will communicate this and set it up in your environment if desired.
Model Overview Information
Outside the admin section, there is also an overview for employees and users of all available models. You can access this by clicking Preferences and then "Information".