# Choose your Chirp's AI model

> Pick the AI model that writes a Chirp's answers, and set its temperature and reply length.

Source: https://www.chirpbuilder.com/help/openrouter/choose-a-model  
Plans: Essentials, Creator, Business, Academy  
Last checked: 2026-10-26

When you're done, your Chirp answers with the AI model you picked.

**Who can do this:** anyone who can edit the Chirp, on any plan. Every plan can use any model.

**What you'll need:** a Chirp, and your OpenRouter key connected so you can try the new model.

## Choose a model

1. Open the Chirp (**Chirps**, then **Edit**) and scroll to **5. Model**.
2. Click the box that shows the current model.

   ![The Model section with Anthropic: Claude Haiku 5.5 chosen, then Let visitors switch models, Temperature 0.7 and Max reply tokens 4000.](https://mjmtbasmrtksluvlnema.supabase.co/storage/v1/object/public/help-media/web-optimized/openrouter/choose-a-model/SS-OR-020-01__model-section__essentials__detail__v01.webp?v=ccc1d31f2d)

3. Find a model: type in **Search models**, click a company such as **Anthropic** or **Google**, or
   change **Sort** to **Cheapest** or **Most capable**. Then click the model.

   ![The model list open: a search box, filters by company, sort options, and the Recommended models with their price per reply.](https://mjmtbasmrtksluvlnema.supabase.co/storage/v1/object/public/help-media/web-optimized/openrouter/choose-a-model/SS-OR-020-02__model-picker__essentials__desktop__v01.webp?v=d72412ab30)

4. Click **Save**. If the Chirp is already published, click **Publish changes** in the bar at the top
   so visitors get the new model too.

## Reading the list

- **Recommended** are good, low-cost models for most Chirps. **Premium** are stronger models that cost
  more per answer. **All models** is everything else OpenRouter offers.
- Under each name: the model's ID, how much text it can read at once (its context), its price per
  million tokens in and out, and about what a typical reply costs, such as **under $0.001 per reply**.
- Labels show what a model can do: **Web search** (it can use web search, images and Actions),
  **Images in** (it can read pictures visitors attach), **Makes images**, **Free** and **Premium
  price**.

If you pick a premium model, a note tells you about what a typical reply costs. Answers are charged to
your OpenRouter account, so consider a credit limit:
[Keep your AI spending in check](https://www.chirpbuilder.com/help/openrouter/spending-limits).

## Temperature

**Temperature** sets how varied the wording is, from 0 to 2. Lower is more focused and consistent;
higher is more creative. 0.7 suits most Chirps. For factual helpers, try 0.3.

## Reply length

**Max reply tokens** is the longest answer the Chirp can give, counted in tokens (about four letters
each). It starts at 4000, which is plenty for most answers. A lower number keeps replies short and a
little cheaper; too low and answers stop mid-sentence.

## If it doesn't work

- **A warning says the model can't use web search or Actions.** Choose a model with the **Web search**
  label, or turn those features off.
- **A warning says the model was removed from OpenRouter.** Click the button it offers to switch to a
  similar model, then save and publish.
- **Answers didn't change for visitors.** You saved but didn't publish: click **Publish changes**.