Choosing the AI model for your AI agent

The instructions below are for desktops and laptops only.

You can choose which AI model your AI agent uses to generate replies. Each model has a different credit cost per reply, so you can select a model that suits your needs and budget.

New AI agents use Auto by default. Auto uses 3 credits per reply. You can switch to a specific model, including models that use fewer credits per reply.

Before you start

Ensure you have enabled AI Assist and created an AI agent. 

To learn more about setting up AI Assist and configuring your AI agent, see this guide:
Getting started with AI Assist

Choosing an AI model

Only a Property Admin can change an AI agent’s model.

To select or change the AI model:

1. Log in to your tawk.to account, and select the correct property.

2. Click Automations on the left navigation bar.

3. Click Agents in the left submenu.

4. Select your AI agent.


5. On the Instructions page, click the Completion Model dropdown.

6. Select your preferred model. Each model in the dropdown displays its credit cost per reply. Review the cost before selecting a model.

Your selection takes effect immediately, including in conversations that are already in progress.

Use the widget preview on the right to test your AI agent’s responses.

Understanding how AI models and credit costs work

Every AI model in your dashboard has a specific credit cost per reply. When your AI agent sends a response, it uses credits based on its assigned model:

  • Auto (Default): All new AI agents run on Auto by default. Auto uses 3 credits per reply. Existing agents keep their current model unless you change them manually.

  • Claude Haiku 4.5 (Recommended): Uses 1 credit per reply and provides fast, accurate, and cost-effective responses for most workflows.

  • 1-credit models: Options like Claude Haiku 4.5, GPT-4o mini, and Gemini Flash cost 1 credit per reply.

  • Higher-credit models: Advanced options range from 2 up to 10+ credits per reply.

Each model shows its credit cost per reply directly inside the selection dropdown menu.

Note: Changing the model does not change your AI Assist subscription plan or billing. The feature is available on all AI Assist plans, and selecting a different model does not change your plan price.

Viewing model usage in Reporting

Reporting shows how many AI Assist credits each model has used during a selected timeframe.

To view model usage:

1. Click Reporting in the left navigation bar.

2. Click Credits under AI Assist in the left submenu.

3. Review the credit usage breakdown by model. Use the timeframe selector at the top right to select the period you want to review.

Note: The Legacy category represents credits used before model-level tracking was introduced, when we did not record which model generated each reply.

Additional considerations

  • Failed replies
    If an AI agent fails to send a reply due to a system or connection issue, you are not charged any credits for that attempt.

  • AI Commands
    Running AI Commands always costs 1 credit, regardless of which model your agent uses.

  • Model availability
    The models shown in the Completion Model dropdown may change over time.

  • Review credit usage
    Check Reporting regularly to understand how many credits each model uses and choose a model that fits your usage needs.

Related guides


If you have feedback about this article, or if you need more help:

Was this article helpful?

1 out of 3 liked this article

Still need help? Message Us