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

# Model configuration

> Do I need to configure a model, and how do I use my own API key instead?

You do not need to configure a model: the free deliveries and both paid plans include model usage. You can bring your own key if you would rather pay your provider directly, or if you want delivery to continue after a paid plan's monthly allowance runs out.

By default, deliveries run on Orbi's platform model channel. That is why you can install, connect and complete a delivery without entering any key at all.

Configuring your own key (**BYOK**, bring your own key) adds your provider account as a model source for your account. Which source a delivery uses depends on your plan:

* **Free plan:** once a key is saved, deliveries run on your key.
* **Paid plan:** deliveries use the monthly allowance included in your plan first. When that allowance is used up, they switch to your key instead of pausing.

## When you should configure a key

| Situation | What to do |
| - | - |
| You just want it to work | Nothing. The platform channel is used automatically |
| You want a specific model, or your compliance rules require your own provider account | Configure BYOK. On a paid plan, your key takes over only after the monthly allowance is used up |
| You have hit the monthly token quota | Configure BYOK. Delivery resumes as soon as you save, and your own key is never paused by the quota |

## Configure it

Click your avatar, choose **Settings**, and open the **AI model** section. The same page also contains message language and notifications. When your monthly allowance is exhausted, **Use your own API key** on the status page links to this section.

1. **Provider**: pick from the list. Selecting one fills in the base URL for you.
2. **baseUrl**: filled automatically, and editable. Change it only if you use a proxy or a private gateway.
3. **model\_id**: choose from that provider's known models, or choose **Enter another model** and type a model ID that is not in the list.
4. **API key**: your key from that provider.

Click **Test connection** first if you want to check the combination. It sends a small request with what you entered and reports either **Connection successful** with the model and how long it took, or the provider's error. Then click **Save model configuration**.

Saving runs the same connection test before anything is stored. If the test fails, nothing is saved: the page shows the error, keeps what you entered and asks you to enter the API key again. A provider outside the built-in list cannot be tested this way and is saved as entered.

After a successful save the page confirms **Model configuration saved**. Saving takes effect for deliveries that start afterwards. If preparing the work environment had failed, saving retries it automatically.

## How your key is stored

Your API key is encrypted at rest and is only ever displayed back to you masked: the field label shows something like **API key (current: ••••1234)**, never the key itself. It is never written to logs.

Do not send your key over email or paste it into a GitHub Issue, pull request or comment. Enter it on this page; that is the only channel you need.

## Supported providers

Selecting one of these fills in its base URL automatically. The "In the dropdown" column is the name the English interface shows; each option also shows its base URL in brackets.

| Provider | In the dropdown | Base URL |
| - | - | - |
| OpenAI | OpenAI | `https://api.openai.com/v1` |
| DeepSeek | DeepSeek | `https://api.deepseek.com/v1` |
| Z.ai (GLM) | Zhipu z.ai | `https://api.z.ai/api/paas/v4` |
| Z.ai Coding Plan | Zhipu z.ai Coding Plan | `https://api.z.ai/api/coding/paas/v4` |
| Zhipu BigModel | Zhipu bigmodel | `https://open.bigmodel.cn/api/paas/v4` |
| Moonshot Kimi (China) | Kimi China | `https://api.moonshot.cn/v1` |
| Moonshot Kimi (International) | Kimi International | `https://api.moonshot.ai/v1` |
| OpenRouter | OpenRouter | `https://openrouter.ai/api/v1` |
| Groq | Groq | `https://api.groq.com/openai/v1` |
| Alibaba DashScope (Qwen) | Qwen (DashScope) | `https://dashscope.aliyuncs.com/compatible-mode/v1` |
| xAI | xAI | `https://api.x.ai/v1` |
| Mistral | Mistral | `https://api.mistral.ai/v1` |
| MiniMax (China) | MiniMax China | `https://api.minimax.chat/v1` |
| MiniMax (International) | MiniMax International | `https://api.minimaxi.com/v1` |
| SiliconFlow | SiliconFlow | `https://api.siliconflow.cn/v1` |
| Together | Together | `https://api.together.xyz/v1` |
| Cerebras | Cerebras | `https://api.cerebras.ai/v1` |
| Volcengine Ark | Volcengine Ark | `https://ark.cn-beijing.volces.com/api/v3` |
| Anthropic (compatibility layer) | Anthropic (compatibility) | `https://api.anthropic.com/v1` |
| Google Gemini | Google Gemini | `https://generativelanguage.googleapis.com/v1beta` |

Google Gemini uses its native API. Previously saved official Gemini compatibility URLs are adapted automatically; you do not need to save them again. Anthropic remains a compatibility option.

Each provider offers a list of known model IDs in the **model\_id** dropdown. That list is a convenience, not a limit: **Enter another model** accepts any model ID the provider serves.

> [!NOTE]
>
> A provider being listed is not a promise that your account has quota, access or rate limit for a given model on it. If a model is not available to your key, the connection test fails, or the delivery fails with the provider's own error.

## Using a provider that is not listed

Choose any provider as a starting point, then edit **baseUrl** to your gateway's URL and type the model ID with **Enter another model**. Any OpenAI-compatible endpoint works. Saving still tests the connection against the URL you entered. Your configuration is stored as you entered it and shown back to you unchanged.

A saved provider that is not in the built-in list, is shown with **(not in the built-in provider list; saved as entered)**, and a saved model that is not in the provider's list is shown with **(saved)**.

## Removing your key

The **AI model** section has no delete button. To go back to the platform channel only, [contact support](https://cloud-docs.orbi.build/troubleshooting.md#getting-help) and ask for your key to be removed. To switch to a different key, just save the new one over it.

## What changes when you use BYOK

| | Platform channel (default) | Your own key |
| - | - | - |
| Who pays for tokens | Included in your plan | You pay your provider |
| When it is used | Free plan: when no key is saved. Paid plan: until the monthly allowance is used up | Free plan: every delivery. Paid plan: after the monthly allowance is used up |
| Monthly token allowance | Paid plans: your plan's allowance (Solo 400,000,000, Pro 1,200,000,000) | Not applied: your usage is measured but never paused |
| Model choice | Operated by Orbi | Yours |
| Usage shown on the status page | **Monthly AI usage**, against the allowance on a paid plan | **Monthly AI usage (your API key)**, on its own line with no limit, once your key has been used |

See [limits and quotas](https://cloud-docs.orbi.build/limits-and-quotas.md) for how the quota is counted.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.