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

# Models

> Browse available LLM models and enable them in the AI Gateway. Filter by provider, capability, and pricing to find the right model.

See [HTTP URLs and authentication](/docs/reference/base-urls.md) and the API reference for complete request and response formats.

## Enabling new Models

To see available Models and enable them for use, navigate to the **Models** page in **AI Gateway**.

Each model row carries the following columns. **Status** and **Model Access** are visible to workspace admins only.

| Column | Description |
| - | - |
| **Name** | The model's display name, provider logo, and model ID. A **New** badge marks models added in the last 14 days, a **Deprecated soon** badge marks models with a scheduled removal date, and a key icon marks providers with a custom API key configured. |
| **Status** | Toggle to enable or disable the model for the **AI Gateway**. |
| **Model Access** | The projects that can use the model: **All Projects**, **No projects** (custom sharing with none selected), or the selected project names. See [Restrict Model Access by Project](#restrict-model-access-by-project). |
| **Price** | Input and output cost per 1M tokens. Image models show a per-image cost instead. Hover the cell for the full pricing breakdown. |
| **Intelligence** | The model's Intelligence Index, the value the [**Smart Router**](/docs/ai-gateway/smart-router.md) bands read from. |
| **Capabilities** | Icons for the modalities and capabilities the model supports. Hover the cell for the full list. |
| **Released** | The model's release date. |
| **Max Output Tokens** | Maximum tokens the model can generate per response. |
| **Context Window** | Total token window (input + output). |
| **Owner** | **Public** for models provided by **Toqan Router**, **Private** for models onboarded to the workspace. |
| **Location** | The region the model is served from. |

Use <kbd> Sort: Newest</kbd> to reorder by **Newest**, **Pricing: Low to High**, **Pricing: High to Low**, **Context: Low to High**, **Context: High to Low**, **Intelligence: High to Low**, **Intelligence: Low to High**, or **Max Output Tokens**. Use <kbd> Columns</kbd> to show or hide individual columns.

> [!TIP]
>
> Use the **Status** toggle to **Enable** a model for use with the **AI Gateway**.

### Row actions

Hover a model's **Name** cell to reveal that row's actions:

| Action | Description |
| - | - |
|  **Copy Model ID** | Copies the model ID to the clipboard, with the workspace prefix applied to private models. |
|  **View Code** | Opens pre-filled cURL, Node.js, and Python snippets for calling the model. |
|  **Access control** | Opens the [project access settings](#restrict-model-access-by-project) for the model. Shown on enabled models. |

Private models carry a **Delete** action in the row menu. Models onboarded through providers that support editing, such as **OpenAI Compatible**, **Azure**, and **Google**, also show **Edit**, which reopens the provider wizard.

### Filters

Use the modality tabs at the top of the list to scope models by type:

<kbd> All</kbd> <kbd> Text</kbd> <kbd> Image</kbd> <kbd> Audio</kbd> <kbd> Speech</kbd> <kbd> Embedding</kbd> <kbd> Moderation</kbd> <kbd> Rerank</kbd> <kbd> Decisions</kbd>

The sidebar provides additional filters:

| Filter | Description |
| - | - |
| **Location** | Filter by region: Europe, United States, Global, APAC, Australia, Singapore |
| **Access** | Toggle **Zero data retention** for ZDR-compliant providers (off by default), or [**BYOK**](/docs/ai-gateway/providers-overview.md) for providers where an API key has been added |
| **Providers** | Filter by LLM provider. See [Providers](/docs/ai-gateway/providers-overview.md) to configure API keys |
| **Status** | Show Enabled or Disabled models. Visible to workspace admins only |
| **Features** | Filter by capability: Base64, Decisions, Code Execution, Image Edit, JSON Mode, PDF, Reasoning, Streaming, Tool Calling, URL, Vision, Web Search |
| **Context length** | Drag the range slider to filter by context window size. The bounds follow the models in the catalog |
| **Intelligence index** | Drag the range slider to filter by Intelligence Index |
| **Input price / 1M tokens** | Drag the range slider to filter by input cost per 1M tokens |
| **Owner** | Filter between Public (**Toqan Router**-provided) and Private (onboarded) models |

To enable a model, toggle it on. It will immediately be available to call with the **AI Gateway**.

### Enable or Disable Models via the API

Models can also be enabled and disabled programmatically using the **Models** API. This is useful for CI/CD pipelines, automation scripts, or infrastructure-as-code workflows.

These endpoints sit on the management plane and authenticate with a Management Key that has `workspace-model` write access. A standard API Key cannot be granted the `workspace-model` domain and is rejected with **403**.

Enable a model:

```bash cURL theme={"theme":{"light":"github-light","dark":"github-dark"}}
curl --request POST \
  --url https://router.toqan.ai/eu/v2/workspace-models \
  --header "Authorization: Bearer $TOQAN_MANAGEMENT_KEY" \
  --header "Content-Type: application/json" \
  --data '{ "model_id": "openai/gpt-5.6-sol" }'
```

Disable a model:

```bash cURL theme={"theme":{"light":"github-light","dark":"github-dark"}}
curl --request DELETE \
  --url https://router.toqan.ai/eu/v2/workspace-models/openai/gpt-5.6-sol \
  --header "Authorization: Bearer $TOQAN_MANAGEMENT_KEY"
```

Both endpoints return **204** on success. Re-enabling an already-enabled model or disabling an already-disabled model is idempotent and also returns **204**.

When **Enforce enabled models** is turned on in General Settings, only models enabled through the dashboard or this API are available for routing. Requests that reference a non-enabled model are rejected.

For full request and response schemas, see [Enable model for workspace](/docs/reference/models/enable-model-for-workspace.md) and [Disable model for workspace](/docs/reference/models/disable-model-for-workspace.md).

### Restrict Model Access by Project

Once a model is enabled, workspace admins can see an  **Access control** icon next to it. Select it to choose which projects can use the model, in one of two modes:

* **All projects** (default): every project in the workspace can use the model.
* **Custom**: each project gets its own on/off toggle, letting admins grant or revoke access per project.

Changes save immediately; there is no separate save action.

> [!NOTE]
>
> Access control is admin-only. Members with other roles see only the models an admin has approved within their chosen projects.

## Onboarding Private Models

Onboard private models by choosing <kbd> Model</kbd> at the top-right of the screen. This is useful when hosting a fine-tuned model or any model deployed on a private provider such as an OpenAI-compatible endpoint, **Azure AI Foundry**, or **AWS Bedrock**.

### Private Models Providers

**OpenAI Compatible**

In the **AI Gateway** sidebar, go to **Models**, then click <kbd> Model</kbd> at the top-right and select **OpenAI Compatible**.

**1. Set the model type**

Choose **Chat Completion**, **Completion**, **Embedding**, or **Image**.

**2. Connect the endpoint**

Enter the **Base URL** of the OpenAI-compatible endpoint, the **Model ID** the provider defines, and an **API Key**.

**3. Identify the model**

Set the **Model name** the gateway exposes, then fill in the token limits, per-token costs, and capabilities.

For the full field reference, see [OpenAI-compatible models](/docs/ai-studio/integrations/providers/open-ai-like.md).

**Azure AI Foundry**

From the [**Azure AI Foundry**](https://ai.azure.com/) project homepage, copy the **API key** and one of the two endpoints shown at the top. **Toqan Router** accepts the following endpoints:

| Endpoint type | Format | What it imports |
| - | - | - |
| Azure OpenAI endpoint | `https://<resource>.openai.azure.com/openai/v1` | OpenAI-compatible deployments |
| Project endpoint | `https://<resource>.services.ai.azure.com/api/projects/<project>` | All deployments from publishers in the project (Anthropic, Cohere, and xAI) |

Paste the endpoint URL exactly as shown in **Azure AI Foundry**. **Toqan Router** does not append any path suffix.
\[!NOTE] To learn more about the Azure AI Foundry deployment, see our [Provider Documentation](/docs/ai-studio/integrations/providers/azure.md).

**1. Open Add Model**

In the **AI Gateway** sidebar, go to **Models**, then click <kbd> Model</kbd> at the top-right and select **Azure**.

**2. Enter credentials**

Enter the **Base URL** and **API Key**.

**3. Fetch deployments**

Click **Fetch deployments** to automatically import all available deployments. The imported models appear in the **Models** list. Toggle each model **Enabled** before use. Enabled models are available for routing requests through [Routing Rules](/docs/ai-gateway/configuration/routing-rules.md).

Reference imported models in code using `<workspacename>@azure/<modelname>` (see [Referencing Private Models in Code](#referencing-private-models-in-code)).

**AWS Bedrock**

To add Bedrock models, first create an [Integration](/docs/ai-studio/integrations/providers/aws-bedrock.md) for the AWS account.

**1. Open Add Model**

In the **AI Gateway** sidebar, go to **Models**, click <kbd> Model</kbd> at the top-right and select **AWS Bedrock**.

**2. Choose the model**

Select the **AWS Integration**, enter the **Inference Profile ARN**, and confirm the **Region** parsed from the ARN.

**3. Validate the connection**

Click **Validate Connection**, then enter a display name and the model details.

**LiteLLM**

To import LiteLLM models, first create an [Integration](/docs/ai-studio/integrations/providers/litellm.md) for the LiteLLM instance. After creation, return to the **AI Gateway** and import models from the connected instance.

### Referencing Private Models in Code

When referencing private models through the SDKs, API, or [Supported Libraries](/docs/ai-studio/integrations/frameworks/overview.md), the model is referenced by the following string: `<workspacename>@<provider>/<modelname>`.

> Example: `corp@azure/gpt-5.6-sol`

## Bring Your Own Key (BYOK)

To start using models, connect provider API keys via [**BYOK**](/docs/ai-gateway/providers-overview.md) in the **AI Gateway** sidebar.


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