---
title: "AI Models"
description: "Choose the AI model that powers your agent and understand how the model affects answer quality, speed, and credit usage."
---

The **AI model** is the engine that writes your agent's replies. You choose it when you create an agent, and you can change it at any time.

## Change the model

- In **Settings**, in the **AI** section, under **AI model**.
- In the **Playground**, under **Model**, so you can compare answers before saving.

## Available models

| Model | Description | Plan | Reads images and PDFs |
| --- | --- | --- | --- |
| **GPT-6 Luna** | Fast and efficient for everyday conversations. The default for new agents. | All plans | No |
| **GPT-5 Nano** | Fastest and most cost-efficient GPT-5 option. | Starter and above | Yes |
| **GPT-5 Mini** | Balanced capability and speed for everyday use. | Starter and above | Yes |

The model picker shows the models available to your team. Some models require a higher plan; they're marked in the picker, with an **Upgrade** link.

Every model accepts voice messages, since they're transcribed first. See [Attachments](/docs/agents/attachments#ai-model-support).

:::note
GPT-4o and GPT-4o Mini are no longer available. Agents that used them were moved to GPT-6 Luna automatically. If one of those agents accepted images or PDFs, choose GPT-5 Nano or GPT-5 Mini to keep them.
:::

:::note
We add and update models over time. The model picker always shows the current list.
:::

## Choosing a model

- **Start with a fast model.** For most support and sales conversations, a fast model gives good answers quickly.
- **Try a more capable model** if your agent needs to follow complex instructions or reason through longer questions.
- **Compare in the Playground.** Ask the same questions with different models before switching.

## Models and credits

Each reply uses [credits](/docs/billing/understanding-credits). More capable models use more credits per reply, so a faster model lets your credits go further.