> For the complete documentation index, see [llms.txt](https://docs.wonderchat.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.wonderchat.io/setup-guides/setting-up-your-chatbot/selecting-between-ai-models-for-your-chatbots.md).

# Selecting Between AI models for your chatbots

A guide to choosing the AI model that powers your agent, what each one costs, and which plans can use it.

Wonderchat offers models from OpenAI, Google, Anthropic and DeepSeek. Each model has its own balance of speed, cost and reasoning. This guide helps you pick the right one.

***

## Where to change the model

1. Go to **Agents** and click **Edit** on your agent.
2. Open the **Model & Behavior** tab.
3. Under **Model**, open the **AI model** menu and pick a model. Models are grouped by provider, and each one shows how many messages a reply uses.

<figure><img src="https://3760420285-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F0WUvfmkdMPXz3x7NXP8i%2Fuploads%2Fgit-blob-d50ac6701cfb8bc84c65df7c8469924dc7c0681b%2Fai-model-picker.png?alt=media" alt="The AI model menu open on the Model and Behavior tab, grouped by provider"><figcaption><p>Model &#x26; Behavior → AI model</p></figcaption></figure>

New agents use **gpt-6-luna**. It's fast, uses 1 message per reply, and is the right choice for most agents.

## Which plans can use which models

* **Fast models** (1 or 2 messages per reply) are on every plan, including Free.
* **Frontier models** (5 or 10 messages per reply) need the **Basic** plan or higher.
* **Claude models hosted in the EU** need the **Scale** plan or higher.

A model your plan can't use is shown greyed out with the plan it needs, for example **Basic+**.

## Message costs and answer depth

The number in each model's name is the messages one reply uses at the default **Answer depth** of **Normal**. **Answer depth**, under the AI model menu, sets how much of your knowledge base the agent reads per answer. **Extended** and **Comprehensive** read more and cost more per reply. The cost is shown under the setting as you change it.

| Model cost  | Normal | Extended | Comprehensive |
| ----------- | :----: | :------: | :-----------: |
| 1 message   |    1   |     5    |       10      |
| 2 messages  |    2   |     6    |       12      |
| 5 messages  |    5   |    10    |       20      |
| 10 messages |   10   |    20    |       40      |

See [Wonderchat Message Credits – Complete Guide](https://docs.wonderchat.io/setup-guides/wonderchat-message-credits-complete-guide.md) for how messages are counted.

***

## OpenAI models

| Model             | Messages per reply | Plan     | Best for                                                                                    |
| ----------------- | :----------------: | -------- | ------------------------------------------------------------------------------------------- |
| **gpt-6-luna**    |          1         | Any plan | The default. Fast, affordable and great for everyday support conversations.                 |
| **gpt-6.1-terra** |          5         | Basic+   | Stronger reasoning for nuanced or multi-part questions, at a fraction of the flagship cost. |
| **gpt-5.6-sol**   |         10         | Basic+   | OpenAI's top-tier reasoning for complex, multi-step and agentic tasks.                      |
| **gpt-5.5**       |         10         | Basic+   | High reasoning for complex, multi-step tasks.                                               |
| **gpt-4o**        |          5         | Basic+   | An older model that's strong at multilingual conversations.                                 |

**In summary:** start with **gpt-6-luna**. Move to **gpt-6.1-terra** if answers need more reasoning, and to **gpt-5.6-sol** for the hardest questions.

## Anthropic Claude models

| Model                | Messages per reply | Plan     | Best for                                                                      |
| -------------------- | :----------------: | -------- | ----------------------------------------------------------------------------- |
| **Claude Haiku 5.5** |          2         | Any plan | Fast, affordable answers from a simple knowledge base.                        |
| **Claude Sonnet 5**  |          5         | Basic+   | Near-Opus quality with balanced speed and cost. The recommended Claude model. |
| **Claude Opus 5.5**  |         10         | Basic+   | The most capable Claude model, for complex research and high-stakes answers.  |

Some teams simply find Claude handles their knowledge base and tone better. If you're unsure, try both and compare.

### Claude models hosted in the EU

If your data must stay in the EU, choose a model from the **Anthropic Claude · EU** group. These models answer from servers in the EU only. They need the **Scale** plan or higher.

| Model                      | Messages per reply |
| -------------------------- | :----------------: |
| **Claude Haiku 4.5 · EU**  |          2         |
| **Claude Sonnet 4.6 · EU** |          5         |
| **Claude Opus 4.6 · EU**   |         10         |

## Google Gemini models

| Model                     | Messages per reply | Plan     | Best for                                                                  |
| ------------------------- | :----------------: | -------- | ------------------------------------------------------------------------- |
| **Gemini 3.1 Flash Lite** |          1         | Any plan | The cheapest, fastest Gemini model, for high-volume FAQ agents.           |
| **Gemini 3.5 Flash**      |          2         | Any plan | Google's newest model. Frontier-level intelligence that runs fast.        |
| **Gemini 3.1 Pro**        |          5         | Basic+   | Strong reasoning and a very large context window, for detailed questions. |

## Other models

| Model                   | Messages per reply | Plan     | Best for                                                           |
| ----------------------- | :----------------: | -------- | ------------------------------------------------------------------ |
| **DeepSeek V4.1 Flash** |          1         | Any plan | Very low cost with a large context window, for high-volume agents. |
| **DeepSeek V4 Pro**     |          5         | Basic+   | DeepSeek's most capable model, with strong reasoning.              |

**Enterprise** plans can also use OpenAI models hosted on Microsoft Azure (**azure/gpt-4.1-mini** and **azure/gpt-4o-mini**, 1 message per reply). They only appear in the menu on Enterprise.

## Retired models

Older models are no longer offered in the menu. If your agent already uses one, it keeps working and stays selected until you change it. Some retired models now run on their successor automatically, for example gpt-5.6-luna now runs on gpt-6-luna and Claude Opus 4.8 runs on Claude Opus 5.5.

We will keep updating this page as we add more models. To request a model, share [**feedback here**](https://wonderchat.featurebase.app/).

***

If you have any more questions, feel free to reach out to us at <mark style="color:purple;">**<support@wonderchat.io>**</mark>
