***

title: Model Settings
sidebarTitle: Model Settings
description: 'Configure AI model, language, and speech behavior.'
-----------------------------------------------------------------

Model Settings control how the AI behaves — the model powering responses, language handling, and formatting preferences.

**Location:** Left Sidebar → Agent Settings → Model tab

<Frame caption="Model tab for Single Prompt agents">
  <img src="https://files.buildwithfern.com/smallest-ai.docs.buildwithfern.com/7d5e3b5025b8c1b6840c9ac7e7440a3134535af269ddf66c3e268d040e3fca8d/products/atoms/pages/platform/building-agents/images/model-settings-sp.png" alt="Model settings" />
</Frame>

***

## AI Model

Choose the LLM powering your agent and its primary language.

| Setting       | Description                           |
| ------------- | ------------------------------------- |
| **LLM Model** | The AI model (Electron, GPT-4o, etc.) |
| **Language**  | Primary language for responses        |

You can also set the model in the [Prompt Section](/atoms/atoms-platform/single-prompt-agents/prompt-section/model-selection) dropdown at the top of the editor.

***

## Speech Formatting

When enabled (default: ON), the system automatically formats transcripts for readability — adding punctuation, paragraphs, and proper formatting for dates, times, and numbers.

***

## Language Switching

Enable your agent to switch languages mid-conversation based on what the caller speaks (default: ON).

### Advanced Settings

When Language Switching is enabled, you can fine-tune detection:

| Setting                         | Range | Default | What it does                               |
| ------------------------------- | ----- | ------- | ------------------------------------------ |
| **Minimum Words for Detection** | 1-10  | 2       | How many words before considering a switch |
| **Strong Signal Threshold**     | 0-1   | 0.7     | Confidence level for immediate switch      |
| **Weak Signal Threshold**       | 0-1   | 0.3     | Confidence level for tentative detection   |
| **Consecutive Weak Signals**    | 1-8   | 2       | How many weak signals needed to switch     |

**Understanding the thresholds:**

* **Strong Signal:** Very confident the caller switched → switches immediately
* **Weak Signal:** Somewhat confident → waits for more evidence
* **Higher thresholds** = More certain before switching (fewer false switches)
* **Lower thresholds** = Quicker to switch (more responsive)

For most cases, the defaults work well. Adjust only if you're seeing unwanted switching behavior.

***

## Related

<CardGroup cols={2}>
  <Card title="Voice Settings" icon="volume" href="/atoms/atoms-platform/single-prompt-agents/agent-settings/voice-settings">
    Speech speed, pronunciation, and detection
  </Card>

  <Card title="Prompt Editor" icon="pen" href="/atoms/atoms-platform/single-prompt-agents/prompt-section/writing-prompts">
    Where you write your agent's instructions
  </Card>
</CardGroup>