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

# Models

> Control which AI models are available to your organization.

Gatelit maintains a catalog of models from OpenAI, Anthropic, Google, and Mistral. Each model has a platform default (enabled or disabled), and your organization can override those defaults to match your needs.

## What you see on the Models page

The **Models** page lists every model in the catalog. Each row shows:

* **Provider** — which AI provider the model belongs to (OpenAI, Anthropic, Google, or Mistral)
* **Model name** — the human-readable label and the underlying model ID used in API calls
* **Type** — model type, such as `chat`, `embedding`, `image`, or `audio`
* **Capabilities** — whether the model supports structured output (JSON schema responses) and function/tool calling
* **Pricing** — the per-request cost rate in USD per 1 million input tokens and per 1 million output tokens
* **Enabled** — a toggle showing whether the model is currently active for your org

## Enabling or disabling a model

<Steps>
  <Step title="Open the Models page">
    In the left sidebar, click **Models**.
  </Step>

  <Step title="Find the model">
    Scroll through the list or use your browser's search to locate the model you want to change.
  </Step>

  <Step title="Toggle the model">
    Click the toggle in the **Enabled** column to turn the model on or off for your organization.
  </Step>
</Steps>

Changes take effect immediately. Any new request for that model is accepted or rejected based on the updated setting. In-flight requests are not affected.

<Info>
  Your org's toggle overrides the platform default. If you re-enable a model that was disabled by default, it becomes available to all users in your org. If you disable a model that was enabled by default, requests for it will be rejected until you re-enable it.
</Info>

## Provider keys and model routing

When a request is routed to a provider, Gatelit checks whether your organization has a BYOK provider key configured for that provider under **Provider Keys**.

* If a BYOK key exists for the provider, all requests to that provider use your key.
* If no BYOK key is configured, the request falls back to the shared platform key.

<Note>
  Managing provider keys is separate from toggling models. Disabling a model prevents it from being called regardless of which key is in use.
</Note>