> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://developers.deepgram.com/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://developers.deepgram.com/_mcp/server.

# Requesting SageMaker Quota

> How to check and increase AWS service quotas for SageMaker endpoint instance types (ml.g4dn.2xlarge, ml.g5.2xlarge, ml.g6.2xlarge, ml.g6e.2xlarge, ml.g7.2xlarge, ml.g7e.2xlarge) required by Deepgram deployments.

AWS enforces default [service quotas](https://docs.aws.amazon.com/general/latest/gr/sagemaker.html) on the number of SageMaker endpoint instances you can run per account per region. Before you can deploy Deepgram on Amazon SageMaker, you may need to request a quota increase for the GPU-accelerated instance types that Deepgram requires.

## Instance types used by Deepgram

Deepgram SageMaker deployments use the following instance types. Each maps to a separate service quota.

| Instance type     | GPU                           | Quota name                           |
| ----------------- | ----------------------------- | ------------------------------------ |
| `ml.g4dn.2xlarge` | NVIDIA T4                     | `ml.g4dn.2xlarge for endpoint usage` |
| `ml.g5.2xlarge`   | NVIDIA A10G                   | `ml.g5.2xlarge for endpoint usage`   |
| `ml.g6.2xlarge`   | NVIDIA L4                     | `ml.g6.2xlarge for endpoint usage`   |
| `ml.g6e.2xlarge`  | NVIDIA L40S                   | `ml.g6e.2xlarge for endpoint usage`  |
| `ml.g7.2xlarge`   | NVIDIA RTX PRO 4500 Blackwell | `ml.g7.2xlarge for endpoint usage`   |
| `ml.g7e.2xlarge`  | NVIDIA RTX PRO 6000 Blackwell | `ml.g7e.2xlarge for endpoint usage`  |

> **Info**
>
> Quota values represent the maximum number of instances of that type you can run simultaneously across all SageMaker endpoints in a single AWS region. A quota of `0` means you cannot deploy that instance type until you request an increase.

## Check your current quota

Before requesting an increase, check the quota you already have.

#### AWS Management Console

#### Open Service Quotas

Sign in to the [AWS Management Console](https://console.aws.amazon.com/) and navigate to [Service Quotas](https://console.aws.amazon.com/servicequotas/home).

#### Select Amazon SageMaker

In the left-hand menu, select **AWS services**, then search for and select **Amazon SageMaker**.

#### Search for the instance quota

In the search bar, enter the instance type you want to check (for example, `g5.2xlarge`). *Find the item named similar to ml.g5.2xlarge for endpoint usage.*

#### AWS CLI

Run the following command, replacing the `--query-text` value with the instance type you want to check:

```bash
aws service-quotas list-service-quotas \
  --service-code sagemaker \
  --query "Quotas[?contains(QuotaName, 'ml.g5.2xlarge') && contains(QuotaName, 'endpoint')]" \
  --output table
```

To check all supported instance types at once:

```bash
for INSTANCE in g4dn.2xlarge g5.2xlarge g6.2xlarge g6e.2xlarge g7.2xlarge g7e.2xlarge; do
  echo "=== ml.$INSTANCE ==="
  aws service-quotas list-service-quotas \
    --service-code sagemaker \
    --query "Quotas[?contains(QuotaName, 'ml.$INSTANCE') && contains(QuotaName, 'endpoint')].{Name:QuotaName,Value:Value}" \
    --output table
done
```

## Request a quota increase

If your current quota is `0` or too low for your deployment, submit a quota increase request.

#### AWS Management Console

#### Open the quota detail page

In the [Service Quotas console for Amazon SageMaker](https://console.aws.amazon.com/servicequotas/home/services/sagemaker/quotas), search for the instance type (for example, `g5.2xlarge`) and select the quota named **ml.g5.2xlarge for endpoint usage**.

#### Request an increase

Select **Request increase at account level**.

#### Enter the new quota value

In the **Increase quota value** field, enter the number of instances you need. For example, enter `4` if you plan to run up to four `ml.g5.2xlarge` endpoint instances in this region.

#### Submit the request

Select **Request**. AWS reviews most SageMaker quota requests within a few hours, though some may take several business days.

#### Repeat for each instance type

If you need quota for additional instance types (`ml.g4dn.2xlarge`, `ml.g6.2xlarge`, `ml.g6e.2xlarge`, `ml.g7.2xlarge`, `ml.g7e.2xlarge`), repeat these steps for each.

#### AWS CLI

Use `request-service-quota-increase` to submit a request. You need the **quota code** for each instance type.

First, look up the quota code:

```bash
aws service-quotas list-service-quotas \
  --service-code sagemaker \
  --query "Quotas[?contains(QuotaName, 'ml.g5.2xlarge') && contains(QuotaName, 'endpoint')].{Code:QuotaCode,Name:QuotaName,Value:Value}" \
  --output table
```

Then submit the increase request using the quota code from the previous command:

```bash
aws service-quotas request-service-quota-increase \
  --service-code sagemaker \
  --quota-code <QUOTA_CODE> \
  --desired-value 4
```

Replace `<QUOTA_CODE>` with the code returned by the lookup command, and `4` with the number of instances you need.

To check the status of your request:

```bash
aws service-quotas list-requested-service-quota-change-history \
  --service-code sagemaker \
  --query "RequestedQuotas[?contains(QuotaName, 'g5.2xlarge')].{Status:Status,Desired:DesiredValue,QuotaName:QuotaName}" \
  --output table
```

## Choosing the right quota value

The quota value determines how many instances of that type you can run simultaneously. Consider the following when choosing a value:

* **Number of Deepgram products** — each product (for example, English STT, Spanish STT, TTS) runs as a separate SageMaker endpoint.
* **Auto-scaling** — if you configure [auto-scaling](/docs/auto-scaling-sagemaker-streaming), set the quota high enough to accommodate the maximum instance count across all endpoints.
* **Multi-region deployments** — quotas are per region. Request increases in every region where you plan to deploy.

> **Warning**
>
> Do not request more instances than you plan to use. Running SageMaker endpoint instances incurs charges for as long as they remain active.

## Troubleshooting

| Symptom                                           | Cause                                                             | Resolution                                                                                                                    |
| ------------------------------------------------- | ----------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------- |
| `ResourceLimitExceeded` when creating an endpoint | The instance quota for the selected type is `0` or fully consumed | Request a quota increase for that instance type                                                                               |
| Quota request remains `PENDING` for several days  | AWS is reviewing the request                                      | Open a support case in the [AWS Support Center](https://console.aws.amazon.com/support/home) referencing the quota request ID |
| Quota increase approved but endpoint still fails  | Quota was increased in a different region                         | Verify the region in the AWS Management Console matches the region where you are creating the endpoint                        |