> ## Documentation Index
> Fetch the complete documentation index at: https://docs.replit.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Explore with different models

> Switch the model behind Replit Design to explore different directions: each model brings its own design sensibility and frontend strengths.

Suggestions explore directions with the same designer; switching models changes the designer. Each model has its own design sensibility and frontend strengths, so the same brief can land somewhere new with a different model behind it.

## Switch the model

Open the mode selector in the chat composer (it shows the current mode, such as **Power**), then select **Primary model** and pick from the list.

<Frame caption="The Agent modes popover with the Primary model list and the Effort slider.">
  <img src="https://mintcdn.com/replit/9Q_lVrxG2Dh35jz1/images/design/replit-design/model-selector.png?fit=max&auto=format&n=9Q_lVrxG2Dh35jz1&q=85&s=012947e6a0df846ce1080dcb1fceaa43" alt="The Agent modes popover with Lite, Economy, and Power modes, Power selected and described as using a higher-performance model for design work, a Primary model list showing Kimi K3, Claude Fable 5 marked Recommended, Claude Opus 5, and GPT-5.6 Sol, and an Effort slider from Low to Max" width="1432" height="786" data-path="images/design/replit-design/model-selector.png" />
</Frame>

The current lineup:

* **Claude Fable 5** (recommended)
* **Claude Opus 5**
* **GPT-5.6 Sol**
* **Kimi K3**

<Note>
  The model lineup changes as new models ship. Check the in-product menu for what your account can use today.
</Note>

Two more controls live in the same popover:

* **Mode** (Lite, Economy, Power): the balance between speed, cost, and capability. Power uses a higher-performance model for design work, best for complex designs and detailed visual changes. See [Agent modes](/features/agent/agent-modes) for the full reference.
* **Effort**: how much reasoning the model spends per response, from Low (fastest) to Max. Each model shows its recommended setting.

## Explore across models

Model switching is another axis of exploration, alongside [suggestions](/design/explore-suggestions):

* **Same brief, different designer**: generate a frame, switch the model, and send the same prompt again. Both directions land as frames on the canvas, so you can compare them side by side.
* **Match the model to the work**: use the recommended model for most work, and switch when a design calls for a different sensibility, or when you want a second opinion on a direction you're unsure about.
* **Switch anytime**: the model applies to your next message, so you can change it mid-project without losing anything on the canvas.

## Next steps

* [Explore](/design/explore): the full exploration toolkit at a glance.
* [Explore suggestions](/design/explore-suggestions): one-click directions proposed from your selected frame.