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

# Models and access

> Discover model capabilities and decide what each project can use.

**Models** shows catalogs from your connected providers, enriched with model details and reference
prices where available. Use it to select a model and manage organization or project blocks.

## Discover and choose a model

1. Select the project whose traffic you want to configure and open **Models**.
2. Search or filter the connected providers' catalogs.
3. Open a model's details to inspect its exact ID, capabilities, limits, and pricing source.
4. Choose an available conversational model with the capabilities your client needs, such as
   function tools for an agent.
5. Test the selected model in [Playground](/gateway/playground) using the intended protocol.

Discovery uses the effective provider connection. A project override therefore determines which
upstream account is queried. Catalogs are cached; refresh after changing provider access. Model
metadata and public prices are reference information, while the provider enforces actual account
permissions, quotas, and model limits.

## Control model access

Owners and Org admins can manage organization blocks from **All projects**. With a project
selected, administrators can add project blocks and inspect inherited restrictions.

Organization and project blocks both apply. Allowing a model at project scope does not remove an
organization block. Project users can inspect and use approved models but cannot change access.
Search and bulk actions apply to the matching catalog entries; review the selection before applying
an access change.

## Bedrock inference profiles

For Bedrock, expand model groups to choose individual inference profiles. Each profile has its own
availability, access setting, and price. Open its details to inspect the exact model ID and
AWS-reported inference regions before selecting it.

The provider connection's region does not guarantee that inference stays in that region. A global
profile may process requests outside Europe. Aixy preserves the selected profile ID and does not
replace a regional profile with a global one. Expanding a model group does not change access;
filters and bulk actions apply to matching entries, including profiles in collapsed groups.

## Use exact model identifiers

A direct ID has the form `provider/model`, for example `openai/gpt-4.1-mini`. The first segment
selects the provider; the rest remains the upstream model ID, including additional slashes.
A [routing model](/guides/routing) uses `aixy/<alias>` and is configured for one project.

A client's authenticated model list contains the conversational models and enabled routes
available to its project. It is a useful check when an editor shows a different list from the
dashboard: verify the editor's key and effective project first.

## Verify a change

Refresh the model list with the project's API key and send a short request. Inspect the effective
provider and model in [Activity](/observe/activity). For an agent, test a function-tool interaction
as well as ordinary text; a successful text reply alone does not establish tool compatibility.


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