# Persona & model

> Give your agent a name, a face, and choose the model that powers its answers.

Source: https://www.lorito.ai/docs/agent/persona-and-model

Your agent's **persona** is how it presents itself to visitors, and its **model** is the AI that generates its replies. Both live in the **Persona** card on the agent page.

## Persona

The persona has two parts:

* **Name.** How the agent refers to itself in conversation (for example, "Aria"). If you leave it blank, the project name is used. This is the name baked into the agent's identity; the label shown in a specific chat header is set per deployment in **Appearance**.
* **Avatar.** The face shown on the chat welcome screen and beside the agent's replies. Upload a square image, or leave it unset to use an automatically generated placeholder.

To change either, open the agent page and click **Edit persona**. Update the name, swap the avatar, and save.

## Model

The model is the AI that reads your knowledge and writes the agent's answers. Every agent has a model setting, and by default it uses the **latest system model**. Lorito keeps this pointed at a strong, balanced model and updates it for you as newer versions become available. You don't have to choose anything to get a good result.

### Choosing a different model

Open **Edit persona** and pick a model from the dropdown. Models are grouped by provider, and each option shows a short hint about its trade-offs. Some are tuned for speed and lower cost, others for higher quality.

What changing the model does:

* It takes effect on the **next message** the agent answers. Existing conversations aren't reprocessed.
* It applies to this agent only. Other agents and projects keep their own settings.
* It does **not** change your knowledge base or system prompt, only which AI generates the replies.

If you ever want to go back, reopen the dropdown and choose **Recommended**. Your agent then follows the latest system model again.

> Different models consume credits at different rates. Faster, lower-cost models stretch your plan's credits further; higher-quality models use more per reply. Pick the balance that fits how your agent is used.
