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

# Model sources: Plus models, your subscriptions, and BYOK

> How the three model sources differ and how to configure them: models included with the Plus plan, your own ChatGPT / Claude and other subscriptions, and BYOK API keys, plus the AI assistant, falling back to Plus models when a key fails, and system task models.

A **model source** determines who provides the model in a session and where usage is counted. AIHubMix Cloud has three model sources: **Plus models** (models included with the Plus plan), **your subscriptions** (your own ChatGPT, Claude, and other subscription accounts), and **BYOK** (API keys you add yourself). All three are configured in **Settings › Personal › Account** and selected in a session through the model picker. See [Harness and models](https://docs.aihubmix.com/en/cloud/models/harness-and-models.md).

## Availability

| Platform | Support | Notes |
| - | - | - |
| Desktop app | Supported | This page describes the desktop app. |
| Web app | Supported | Same as the desktop app. |
| Mobile web | Supported | Settings use the mobile layout. |

## Entry points

* **Settings › Personal › Account**: below the Plus card are three sections, **Your subscriptions**, **BYOK**, and **System task models**.
* Model picker: the pencil icon next to a group title opens the edit window. When a source is empty, **Connect subscription** and **Add API key** open the add window in place.
* **Model sources (subscriptions / BYOK)** in the command menu `⇧⌘P` (`Ctrl+Shift+P` on Windows).

## Three model sources

| | Plus models | Your subscriptions | BYOK |
| - | - | - | - |
| What it is | Models included with the Plus plan | Your own subscription accounts (ChatGPT, Claude, and others) | Your own API keys (gateways, model vendors, or custom services) |
| Where usage is counted | The Plus plan, with no extra charge | The plan limits of that subscription account | The corresponding provider, at its prices |
| How to connect | Available once you get Plus | Sign in and authorize on your cloud server | Enter an API key |
| Location in settings | First row of the **BYOK** section | **Your subscriptions** section | **BYOK** section |
| Model picker tab | BYOK (first group) | Subscriptions | BYOK |
| Compatible Harnesses | As shown in the model picker | By subscription type. See [Supported subscriptions](#supported-subscriptions) | By API format. See [API format and Harness](#api-format-harness) |

AIHubMix is also a regular gateway provider in BYOK: when you add your own AIHubMix API key, it is billed by that provider like other keys, and is unrelated to Plus.

## Plus models

**Plus models** are the models included with the Plus plan. Usage counts toward the plan with no extra charge. Which models are included follows the Plus subscription dialog and what the product shows.

### Where Plus models are

* The first row of the **Settings › Account › BYOK** section: it has the AIHubMix icon and shows **Included with Plus** (**Available with Plus** if you do not have Plus), the model count, and the AI assistant. This row cannot be deleted and does not count toward the key limit.
* The first group in the BYOK tab of the model picker.

### Plus limit status

When you have Plus, the Plus models row shows a status only when something is wrong with the limits. No amounts or usage numbers are shown:

| Status | What you see |
| - | - |
| A period limit is reached | **Daily / Weekly / Monthly usage limit reached; resets time** |
| Usage is used up | **Usage limit reached** |
| Being prepared or temporarily unavailable | **Getting ready, available soon** or **Temporarily unavailable**. Check again later |

### Choose Plus models

You can decide which models appear in the Plus models group:

1. Click the pencil icon on the Plus models row (or in the model picker, hover the Plus models group title and click the pencil icon).
2. In **Your models**, select or remove models. See [Two-column model picker](#model-columns).
3. Click **Save**.

After you make changes, **Reset to default** appears at the bottom. Click it to restore all models in the plan. To add back models you removed earlier, also use **Reset to default** first.

## AI assistant

The **AI assistant** is a background model for each model source, used to handle background tasks outside sessions, such as resolving merge conflicts and organizing context across sessions. The default is **Auto**, which picks a low-cost, low-latency model from that source. Usually you do not need to change it.

To change it, click **AI assistant · model name** on the matching row of the Account page (the Plus models row, a subscription account row, or a key row) and choose one. You can also click **Change** in the window when adding a subscription account or API key.

## Your subscriptions

**Your subscriptions** connects your own subscription accounts to AIHubMix Cloud, and models run on that account's plan limits. You sign in and authorize on your own cloud server, and AIHubMix does not store or relay your subscription credentials.

### Supported subscriptions

Which subscriptions you can add follows the **Add a subscription** window. Each provider shows its connection method and the matching Harness (**Uses X by default**):

| Subscription | Connection method |
| - | - |
| ChatGPT | Browser sign-in, or **Use a device code instead** |
| Claude | Authorization code sign-in |
| OpenCode | Access token |
| GitHub Copilot | Device code sign-in |
| GLM | Paste key |

* The plan tier is determined by the authorized account and does not need to be selected manually.
* When **Add a subscription** is grayed out, adding is not available for now. Connected accounts still work.

### Add a subscription account

Before you start, make sure your cloud server is running. Otherwise you see **Your cloud server isn’t running. Make sure it’s running on the Cloud server page, then connect.**

1. Open **Settings › Account** and click **Add a subscription** in the **Your subscriptions** section (click **+** if you already have accounts). You can also click **Connect subscription** when the Subscriptions tab of the model picker is empty.
2. Choose a provider, confirm the **Name** (used to tell apart multiple accounts from the same provider), and click **Continue**.
3. Read the terms, check **I understand and agree**, and click **Continue**.
4. The window provides an **Authorization URL** that you can **Copy** or **Open**.
5. Sign in and authorize in the browser, then complete the last step for your provider:
   * ChatGPT (browser sign-in): after signing in, the browser goes to a page that does not load. Copy the full address from the address bar, paste it into **Callback URL**, and click **Submit**.
   * ChatGPT (device code): open the link in any browser or on your phone, and enter the **Verification code** from the window. If you cannot sign in, turn on device code sign-in in your ChatGPT account's security settings first.
   * Claude: paste the authorization code from the authorization page into **Authorization code** and click **Submit**.
6. It shows **Connected name (email)**, and the account appears in the list.

| Situation | Message / what to do |
| - | - |
| The account limit for this provider is reached | **You’ve reached the account limit for this provider. Disconnect one before adding another.** |
| The cloud server version is too old | **Your cloud server needs an update before you can add more accounts** |
| Another account from the same provider is already being added | Finish or cancel it first |
| Sign-in cancelled or not completed | Click **Retry** |

### Multiple accounts and account names

You can connect multiple accounts from the same provider, for example one personal and one for work. They are grouped by name in the model picker. Accounts are not switched automatically when limits are reached.

* The default name is the provider name, followed by "ChatGPT (2)", "ChatGPT (3)", and so on.
* Rename: on the Account page, click the pencil icon on the account row. The name cannot be empty, can be up to 32 characters, and must be unique across all subscription accounts (case-insensitive).

### Default account

When you have multiple accounts from the same provider, one of them is marked **Default**, and new sessions using that provider's subscription use it by default. Click **Make default** on another account row to change it.

### Reconnection needed

| Account page shows | What you can do |
| - | - |
| **Authorization expired, this plan can’t be used for now** | Click **Reconnect** |
| **Signed in from a terminal, so your agent can read the credentials. Reconnect here instead.** | Click **Reconnect** to authorize from here instead |
| Other error messages | Follow the prompt |

To reconnect: click **Reconnect**, then **Authorize with name**, and complete steps 4 to 6 of [Add a subscription account](#add-subscription).

### Disconnect a subscription account

On the Account page, click **Disconnect** on the account row. The account is removed immediately, with no confirmation. Sessions using this account pause, and **Subscription account disconnected** opens. You can click **Use another model** or **Later**.

### Plan limits reached

When the subscription account used by a session reaches its limit, the session pauses, finished changes and the environment are kept, and **name usage limit reached** opens, stating which limit was reached and roughly when it resets. AIHubMix Cloud does not switch accounts automatically.

| Situation | What you can do |
| - | - |
| Other accounts from the same provider exist | Choose one in **Account to switch to** and click **Continue with this account**. The model stays the same. |
| No other accounts | Click **Add account** to add one on the Account page, or continue with another model. |
| You want to switch models | Click **Use another model** to open the model picker (the Harness stays the same). The session continues automatically after the switch. |
| Not now | Click **Later** to close the dialog. |

You can view the usage and reset times of each subscription account's 5-hour and 7-day limits under **Plan limits** in **Settings › Usage**. See [Usage and billing records](https://docs.aihubmix.com/en/cloud/settings/usage-and-billing.md).

## BYOK

**BYOK** (Bring Your Own Key) uses API keys you add yourself, billed by the corresponding provider. You can add up to 10 keys (the Plus models row does not count). After 10 keys, **+** is grayed out.

### Supported providers

When adding an API key, you can choose from the following providers. Selecting one fills in the Base URL and API format:

* **Gateways**: AIHubMix, OpenRouter, SiliconFlow, Together AI, Groq, Fireworks AI, Vercel AI Gateway, Cloudflare AI Gateway, Novita AI, Azure OpenAI.
* **Model vendors**: OpenAI, Anthropic, Google Gemini, DeepSeek, xAI, Mistral AI, Moonshot (Kimi), Zhipu GLM, Qwen (Alibaba Bailian), MiniMax, Doubao (Volcano Ark).
* **Other**: Custom, where you enter your own Base URL and API format.

Preset providers default to **OpenAI compatible**, except Anthropic, which defaults to **Anthropic compatible**. For Cloudflare AI Gateway and Azure OpenAI, replace the account, gateway, or resource name in the address with your own.

### Add an API key

1. Open **Settings › Account** and click **+** (**Add API key**) in the **BYOK** section. You can also click **Add API key** when the BYOK tab of the model picker is empty.
2. Search for or choose a provider in **Provider**. Click **Change** if you need to edit the Base URL or API format.
3. Enter the **API key**. Models load automatically, and recommended models are selected automatically. Add or remove models as needed. See [Two-column model picker](#model-columns).
4. **Name (optional)**: if left empty, it is shown as "Vendor ··last 4". Up to 64 characters.
5. Click **Add**. The key appears in the BYOK list.

* If loading models fails (including an invalid key), the reason is shown. You can **Retry**, or type a model name to add it manually.
* **Add** is disabled when no provider is selected, no models are selected, or the Base URL is incomplete.
* After saving, only a masked key is shown.
* When the Base URL is a private network address, sessions running on the cloud server cannot reach this key. It works only in local projects.

### Edit and delete keys

Each key is one row on the Account page, with a pencil (edit) and a trash can (delete) at the end of the row. You can also click the pencil icon next to a group title in the model picker to edit it.

* The provider of a saved key cannot be changed. The Base URL and API format are editable after you click **Change**.
* Save with **API key** empty to keep the original key. Enter a new key and save to replace it.
* Click the trash can to delete the key immediately, with no confirmation.

### Two-column model picker

The key edit window and the Plus models edit window use the same two-column model picker:

* Left column, **All models**: recommended models come first and are marked **Recommended**. The search box filters only the left column. Click a model row to select or deselect it.
* Right column, **Added**: selected models. Click the close icon to remove one.
* In BYOK, you can type a model ID that is not in the list into the search box and click **Add “model ID”** to add it. Each key can have up to 200 models selected.

### API format and Harness

Each key has an API format: **OpenAI compatible** or **Anthropic compatible**. Claude Code needs an Anthropic compatible API, and Codex CLI needs an OpenAI compatible API.

When the format does not match, this key's group in the model picker is grayed out with the reason (for example **API format not compatible with Claude Code**). In a new session, clicking one of these models switches to a compatible Harness automatically. If your service also offers the other API, you can change the Base URL and API format in the edit window, or add another key for the same service.

### Fall back to Plus models when a key fails

To the right of the **BYOK** section title is a toggle, **Fall back to Plus models**. It is off by default and takes effect as soon as you click it.

* Off: if your key fails, the turn stops with an error.
* On: if your key fails (usage limit reached, rate limited, or invalid key), this turn continues on the models included with Plus. If you do not have Plus yet, it takes effect after you subscribe.
* The toggle is disabled until you add a key.

## System task models

The **System task models** section in **Settings › Account** shows, read-only, the models background system tasks (for example **Environment repair** and **Chat titles**) currently use. Each row shows "model name · source · Effort X".

* To change one, change the **AI assistant** on the corresponding subscription row, key row, or Plus models row.
* When it shows **Couldn’t load**, tasks are not affected. You can click **Retry**.

## Notes and limitations

* Accounts are not switched automatically when plan limits are reached or a key fails. Only when **Fall back to Plus models** is on does a key failure switch to Plus models.
* Deleting a key or disconnecting a subscription account has no confirmation and takes effect immediately.
* Adding a subscription requires the cloud server to be running.
* Editing a key or Plus models does not change the model used by sessions that have already started.
* The Plus model list, models available with subscriptions, and recommended models follow what the product shows.

## Related

* [Harness and models](https://docs.aihubmix.com/en/cloud/models/harness-and-models.md)
* [Model FAQ](https://docs.aihubmix.com/en/cloud/models/model-faq.md)
* [Settings › Account](https://docs.aihubmix.com/en/cloud/settings/account.md)
* [Usage and billing records](https://docs.aihubmix.com/en/cloud/settings/usage-and-billing.md)
* [Plus plan](https://docs.aihubmix.com/en/cloud/plus/plus-plan.md)


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