You can use Azure AI Foundry to discover and deploy Voyage AI by MongoDB models in your Azure environment. Azure AI Foundry provides an Azure-native path for model discovery, deployment, billing, and access management.
Using Voyage AI through Azure AI Foundry provides the following benefits:
Data flow and access control. Azure AI Foundry supports private networking and Azure-native access controls so you can align model access with your Azure network and security requirements. For private setups, validate the specific deployment mode and networking configuration that your Foundry project uses in the Azure portal.
Reliability and compliance backed by Azure. Billing, identity, and infrastructure controls are managed through Azure.
Procurement and billing through Azure. You can use your Azure subscription, billing relationship, and applicable credits instead of establishing a separate commercial relationship for model hosting.
Available Models
To see the Voyage AI by MongoDB models currently available in Azure AI Foundry, open the Voyage AI publisher page in Azure AI Foundry.
Model availability can change over time. To learn more about Voyage AI model capabilities and recommended use cases, see Models Overview.
Pricing
Azure AI Foundry pricing depends on the deployment type that Azure uses for the selected Voyage AI model and the region in which you deploy it.
Pricing component | What to expect |
|---|---|
Azure subscription billing | Charges are billed through your Azure subscription. |
Model cost | The Azure AI Foundry deployment UI shows the pricing details for the selected model or deployment option before you deploy. |
Infrastructure cost | Dedicated or managed compute deployments can include hourly infrastructure charges that vary by accelerator SKU and region. |
Region sensitivity | Total cost can vary by Azure region and deployment configuration. |
Azure AI Foundry offers two deployment billing models:
Standard Foundry deployments are typically billed by token usage or provisioned throughput, depending on the deployment type.
Managed compute deployments are billed hourly per accelerator SKU.
The Azure portal or Azure AI Foundry deployment flow is the source of truth for the exact price shown at deployment time. For more information, see Microsoft documentation for deployment options in Azure AI Foundry and the Azure pricing calculator.
Prerequisites
Before you deploy a Voyage AI model through Azure AI Foundry, ensure that you have the following:
An active Azure subscription.
Permissions to create or use Azure AI Foundry resources in the target subscription.
Access to an Azure AI Foundry project, or permission to create one.
Sufficient quota in the target region for the deployment type that Azure AI Foundry requires for the selected Voyage AI model.
An Azure AI hub resource in the same region where you plan to deploy the Voyage AI model through Azure AI Foundry. For setup guidance, see Microsoft documentation for creating hubs and projects.
If your organization requires private networking, an existing virtual network and the Azure permissions needed to create and approve any required private endpoint connections.
Quota
This section includes guidance on how to verify and request quota before deployment.
The quota model in Azure AI Foundry depends on the deployment type:
Some deployment paths use Azure AI Foundry quotas rather than Azure virtual machine quotas.
Other deployment paths might surface infrastructure-oriented sizing and region availability choices in the deployment flow.
Because Azure AI Foundry can route different model families through different deployment modes, verify quota from the Azure portal and the Azure AI Foundry deployment UI before you begin.
Check Quota
For Azure AI Foundry quota guidance, see Microsoft documentation on deployment options.
Request Quota Increase
If you don't have sufficient quota or capacity:
Use the Azure portal to request the relevant quota increase for the required service or region.
If the deployment flow indicates Azure AI Foundry-specific quota requirements, follow the quota workflow linked from the Azure UI.
Retry the deployment after the increase is approved.
Because Azure AI Foundry deployment types evolve, use the Azure portal and deployment workflow as the source of truth for the exact quota process.
Prepare Networking
If your organization does not require private networking, you can use the default Azure AI Foundry networking options that the deployment supports.
If your organization requires private connectivity, prepare your networking before deployment:
Identify the virtual network and subnet that your Azure AI Foundry project or client applications will use.
Confirm that you have permission to create private endpoints and approve private endpoint connections if your organization uses them.
Confirm DNS resolution and routing requirements for private access.
Validate any outbound access policies that your applications, validation hosts, or agent tools require.
For Azure AI Foundry private networking guidance, see Microsoft documentation on network isolation for Azure AI Foundry.
Deploy a Voyage AI Model in Azure AI Foundry
In this section, you deploy a Voyage AI by MongoDB model from Azure AI Foundry.
Because Voyage AI is provided in Azure AI Foundry as a third-party managed compute offering, Azure uses a deployment wizard that provisions GPU-backed infrastructure and deploys the selected Voyage AI model on top of it.
The Azure AI Foundry portal selects the appropriate deployment path for the chosen model. Depending on the model and Azure's current integration for that listing, this can affect the fields you see during deployment and the way you connect to the deployed model.
Deployment Settings
Review the following settings carefully during deployment:
Setting | Guidance |
|---|---|
Subscription | Use the Azure subscription that should receive billing for the deployment. |
Project | Use the Azure AI Foundry project that should own the model deployment. |
Region | Choose a supported region with sufficient capacity and the networking profile your organization requires. |
Deployment type | Use the deployment type offered by Azure for the selected Voyage AI model. |
Authentication | Use the project keys that Azure AI Foundry provides for the deployment. |
Networking | If required, apply private networking or private endpoint settings supported by your Foundry resource and project. |
Confirm Completion
After the deployment succeeds:
Open the deployed model or deployment details page in Azure AI Foundry.
Capture the endpoint information, deployment name, authentication method, and any region or networking details you need for clients.
Record whether the deployment uses the unified Azure AI Foundry project endpoint or another Azure-managed connection pattern exposed by the portal.
Connection Details
Azure AI Foundry connection details depend on the deployment mode that Azure uses for the selected model. Use the deployment details page in Azure AI Foundry as the source of truth.
After deployment, capture the following values from Azure AI Foundry:
Value | Why you need it |
|---|---|
Project endpoint or model endpoint | Your client uses this base URL to send inference requests. |
Deployment name | Some SDKs and API routes require a deployment identifier. |
Authentication method | You need the keys or tokens that Azure AI Foundry provides for the deployment. |
Region | Useful for governance, troubleshooting, and latency expectations. |
Networking constraints | Indicates whether clients must connect from approved networks or through private access. |
Azure AI Foundry commonly exposes one of the following connection patterns:
Unified Foundry project endpoint. Some Azure AI Foundry deployments are invoked through the project's unified endpoint, with the deployment name supplied in the request path or payload.
Private or network-isolated access. If your Foundry project is configured for private access, clients must run from approved networks and resolve Azure private endpoints correctly.
Authentication
Use the authentication method shown in Azure AI Foundry for the deployed model. Voyage AI model deployments currently support key-based authentication only:
- Azure AI Foundry project keys or other key-based credentials that the deployment exposes.
Verify Connectivity
After deployment, verify that your application can connect to the deployed model.
If your deployment uses private networking, also do the following:
Run the test from a client inside the approved network boundary.
Validate DNS resolution to the expected private endpoint, if applicable.
Confirm that firewall, subnet, and private endpoint settings allow the request.
For private networking validation, see Microsoft documentation on network isolation for Azure AI Foundry.
Example Validation Workflow
Use the following workflow to validate the deployment end to end:
Troubleshooting
If you can't connect to the deployment:
Confirm that the model deployment completed successfully.
Confirm that you are using the correct endpoint and deployment name.
Confirm that your authentication method matches the one shown in Azure AI Foundry.
If private networking is enabled, confirm DNS, private endpoint approval, subnet routing, and firewall configuration.
If the deployment failed before completion, review Azure quota, region availability, and any capacity limits shown in the deployment flow.
Delete a Voyage AI Deployment
To avoid ongoing charges, delete model deployments that you no longer need.
If your setup includes private endpoints or other supporting network resources that were created only for this deployment, review whether they should also be removed.