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

# NVIDIA

> Add models from NVIDIA's hosted API to TypingMind as custom models. Set the endpoint, add your NVIDIA API key and route requests via TypingMind Cloud.

You can chat with models from NVIDIA's hosted API in TypingMind. Add each model as a custom model, using your own NVIDIA API key.

> [!NOTE]
>
> NVIDIA's API only works in TypingMind when **Route requests via TypingMind Cloud** is on. Step 4 shows how to turn it on.

## Step 1: Get an NVIDIA API key

1. Sign in at [build.nvidia.com](https://build.nvidia.com).
2. Create an API key and copy it.
3. Open the page of the model you want to use and copy its model ID.

## Step 2: Open the custom model form

1. In TypingMind, click **Models** on the left sidebar.
2. Click **Add custom model** at the top right.
3. Keep the **Create Manually** tab open.

## Step 3: Enter the model details

1. **Name:** a name you will recognize in the model list.
2. **Model ID:** the model ID you copied from NVIDIA.
3. **Context Length:** the context length that NVIDIA lists for the model.
4. On the **Authentication** tab, set:
   * **API Type:** OpenAI Chat Completions API
   * **Endpoint URL:**
     ```text theme={null}
     https://integrate.api.nvidia.com/v1/chat/completions
     ```
   * **Authentication Type:** API Key via HTTP Header
   * **Header Key:** `Authorization`
   * **Header Value:** `Bearer` followed by a space and your NVIDIA API key

![Add Custom Model form with the NVIDIA endpoint URL, API Key via HTTP Header and the Authorization header key](https://mintcdn.com/typingmind/tJQGZV4X5d7eE71d/images/nvidia-custom-model-authentication.jpg?fit=max\&auto=format\&n=tJQGZV4X5d7eE71d\&q=85\&s=41b5bdbb5a8c83bb2780b7cf0439e38b)

## Step 4: Route requests via TypingMind Cloud

1. Open the **Request Options** tab.
2. Under **TypingMind AI Gateway Options**, turn on **Route requests via TypingMind Cloud**.
3. Under **TypingMind AI Gateway Expiration**, keep **Never expires**. If you choose **Custom duration**, routing turns off when the time ends and the model stops working.

This sends requests through TypingMind's server, so your browser doesn't block calls to NVIDIA's API (CORS restrictions).

![Request Options tab with Route requests via TypingMind Cloud turned on and the gateway expiration set to Never expires](https://mintcdn.com/typingmind/tJQGZV4X5d7eE71d/images/nvidia-custom-model-route-typingmind-cloud.jpg?fit=max\&auto=format\&n=tJQGZV4X5d7eE71d\&q=85\&s=de4f8ca3aaf32cf19b32edf1461be725)

## Step 5: Test and save

Click **Test & Save**. When the test passes, the model appears in your model list. Select it in the model selector above the chat input to start chatting.

## FAQ

> [accordion] Why does the test fail with a network or CORS error?
>
> Check that **Route requests via TypingMind Cloud** is on in the **Request Options** tab. NVIDIA's API doesn't work without it.

> [accordion] Why does the test fail with an authentication error?
>
> Check the **Header Value**. It must start with `Bearer` and a space, followed by your full NVIDIA API key.

> [accordion] Can I add more than one NVIDIA model?
>
> Yes. Add each model as its own custom model with its own model ID.

## Related guides

* [Manage models](https://docs.typingmind.com/manage-and-connect-ai-models/manage-ai-models.md)
* [Set up API keys](https://docs.typingmind.com/manage-and-connect-ai-models/set-up-api-keys.md)
* [Google Gemini (Custom)](https://docs.typingmind.com/manage-and-connect-ai-models/google-gemini-as-custom-model.md)
* [Context length](https://docs.typingmind.com/manage-and-connect-ai-models/model-context-length.md)


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