View a markdown version of this page

Responses API on the bedrock-mantle endpoint - Amazon Bedrock
Supported Regions and EndpointsPrerequisitesModels APIResponses API

Responses API on the bedrock-mantle endpoint

Amazon Bedrock provides the OpenAI Responses API on both the bedrock-runtime and bedrock-mantle endpoints. The API lets you use familiar OpenAI SDKs and tools with Amazon Bedrock models, so you can migrate existing applications with minimal code changes—simply update your base URL and API key. For new applications, we recommend the bedrock-runtime endpoint.

The two endpoints don't have identical feature support. Requests on bedrock-runtime are always synchronous, server-side tools aren't available, and only the default project is supported. For the full comparison, see Endpoints supported by Amazon Bedrock, and for the details of each difference, see Using the Responses API on the bedrock-runtime endpoint.

Important

When using the OpenAI SDK with Amazon Bedrock, you must point it to the Amazon Bedrock endpoint, not the OpenAI endpoint. Set the following environment variables, choosing the base URL for the endpoint you want:

# bedrock-runtime (recommended)
OPENAI_BASE_URL="https://bedrock-runtime.<your-region>.amazonaws.com/openai/v1"

# bedrock-mantle
OPENAI_BASE_URL="https://bedrock-mantle.<your-region>.api.aws/v1"

OPENAI_API_KEY="<your Bedrock API key>"

Do not use your OpenAI API key or the OpenAI base URL (https://api.openai.com/v1). Those connect to OpenAI directly, not to Amazon Bedrock. To create a Amazon Bedrock API key, see API keys.

Key benefits include:

  • Asynchronous inference – Support for long-running inference workloads through the Responses API. Available on bedrock-mantle only.

  • Stateful conversation management – Automatically rebuild context without manually passing conversation history with each request

  • Simplified tool use – Streamlined integration for agentic workflows

  • Flexible response modes – Support for both streaming and non-streaming responses

  • Easy migration – Compatible with existing OpenAI SDK codebases

Each endpoint is governed by its own set of quotas. For Responses traffic on bedrock-runtime, the model's tokens-per-minute and tokens-per-day quotas apply, and they're shared with the other inference APIs on that endpoint — see Quotas for the bedrock-runtime endpoint. For bedrock-mantle, see Quotas for the bedrock-mantle endpoint.

Supported Regions and Endpoints

On the bedrock-runtime endpoint, the Responses API is available in every AWS Region where that endpoint is available, including the AWS GovCloud (US) Regions. For the list, see Regional availability by endpoints. Which models support the API on each endpoint is listed in Endpoint availability by models.

The bedrock-mantle endpoint is available in the following AWS Regions:

Region Name Region Endpoint
US East (Ohio) us-east-2 bedrock-mantle.us-east-2.api.aws
US East (N. Virginia) us-east-1 bedrock-mantle.us-east-1.api.aws
US West (Oregon) us-west-2 bedrock-mantle.us-west-2.api.aws
Asia Pacific (Jakarta) ap-southeast-3 bedrock-mantle.ap-southeast-3.api.aws
Asia Pacific (Mumbai) ap-south-1 bedrock-mantle.ap-south-1.api.aws
Asia Pacific (Sydney) ap-southeast-2 bedrock-mantle.ap-southeast-2.api.aws
Asia Pacific (Tokyo) ap-northeast-1 bedrock-mantle.ap-northeast-1.api.aws
Europe (Frankfurt) eu-central-1 bedrock-mantle.eu-central-1.api.aws
Europe (Ireland) eu-west-1 bedrock-mantle.eu-west-1.api.aws
Europe (London) eu-west-2 bedrock-mantle.eu-west-2.api.aws
Europe (Milan) eu-south-1 bedrock-mantle.eu-south-1.api.aws
Europe (Stockholm) eu-north-1 bedrock-mantle.eu-north-1.api.aws
South America (São Paulo) sa-east-1 bedrock-mantle.sa-east-1.api.aws
AWS GovCloud (US-West) us-gov-west-1 bedrock-mantle.us-gov-west-1.api.aws

Prerequisites

Before using OpenAI APIs, make sure you have the following:

  • Authentication – You can authenticate using:

    • Amazon Bedrock API key (required for OpenAI SDK)

    • AWS credentials (supported for HTTP requests)

  • OpenAI SDK (optional) – Install the OpenAI Python SDK if using SDK-based requests.

  • Environment variables – Set the following environment variables:

    • OPENAI_API_KEY – Set to your Amazon Bedrock API key

    • OPENAI_BASE_URL – Set to the Amazon Bedrock endpoint for your region (for example, https://bedrock-runtime.us-east-1.amazonaws.com/openai/v1 or https://bedrock-mantle.us-east-1.api.aws/v1)

  • Permissions – The actions you need depend on the endpoint. On bedrock-mantle, inference authorizes bedrock-mantle:CreateInference. On bedrock-runtime, it authorizes bedrock:InvokeModel on both the inference target and your account's default project, and managing stored responses authorizes bedrock:GetInvoke, bedrock:CancelInvoke, and bedrock:DeleteInvoke on that project. For policy examples, see Prerequisites for running model inference.

Models API

The Models API allows you to discover available models in Amazon Bedrock powered by Mantle. Use this API to retrieve a list of models you can use with the Responses API. For complete API details, see the OpenAI Models documentation.

List available models

To list available models, choose the tab for your preferred method, and then follow the steps:

OpenAI SDK (Python)
# List all available models using the OpenAI SDK
# Requires OPENAI_API_KEY and OPENAI_BASE_URL environment variables

from openai import OpenAI

client = OpenAI()

models = client.models.list()

for model in models.data:
    print(model.id)
HTTP request

Make a GET request to /v1/models:

# List all available models
# Requires OPENAI_API_KEY and OPENAI_BASE_URL environment variables

curl -X GET $OPENAI_BASE_URL/models \
   -H "Authorization: Bearer $OPENAI_API_KEY"

Responses API

The Responses API provides stateful conversation management with support for streaming, background processing, and multi-turn interactions. For complete API details, see the OpenAI Responses documentation.

Note

Not all models support the Responses API. To see which models support the Responses API, see API compatibility by models.

How the Responses API stores conversation state

The Responses API can use stored state to enable multi-turn conversations and let you reference previous turns through the previous_response_id parameter. Storage is enabled by default but can be disabled per request through the store parameter. Stored responses are scoped by Project. A response from one Project cannot be used as the previous response or read in a second Project. For more information about Projects, see Projects (OpenAI-compatible).

  • When store is true (the default), Amazon Bedrock retains the response, including the input and output, for 30 days. During this window you can chain follow-up requests by passing previous_response_id and retrieve the response with GET /v1/responses/{id} on bedrock-mantle, or GET /openai/v1/responses/{id} on bedrock-runtime. After 30 days, the response is automatically deleted and is no longer retrievable.

  • When store is false, Amazon Bedrock does not retain any data from the request or response. The previous_response_id parameter cannot be used to continue the conversation.

The default value is true to match the OpenAI Responses API specification. Customers who do not want Amazon Bedrock to retain conversation data should explicitly set store to false on every request, or set the account's data retention mode to none, which rejects an explicit store=true outright. For more information, see Data retention.

Stored data is encrypted at rest and scoped to the calling AWS account's Project resource. The data is stored solely to service your requests and is not used or retained for any other purpose. On bedrock-mantle, it is kept in the AWS Region the request was sent to. On bedrock-runtime, a request that uses cross-Region inference can be processed in another AWS Region, and the response is stored in the Region that processed it — so a request that uses a global inference profile can store data in any commercial Region that profile routes to. If you have data residency requirements, use a geographic inference profile rather than a global one.

Basic request

To create a response, choose the tab for your preferred method, and then follow the steps:

OpenAI SDK (Python)
# Create a basic response using the OpenAI SDK
# Requires OPENAI_API_KEY and OPENAI_BASE_URL environment variables

from openai import OpenAI

client = OpenAI()

response = client.responses.create(
    model="openai.gpt-oss-120b",
    input=[
        {"role": "user", "content": "Hello! How can you help me today?"}
    ]
)

print(response)
HTTP request

Make a POST request to /v1/responses:

# Create a basic response
# Requires OPENAI_API_KEY and OPENAI_BASE_URL environment variables

curl -X POST $OPENAI_BASE_URL/responses \
   -H "Content-Type: application/json" \
   -H "Authorization: Bearer $OPENAI_API_KEY" \
   -d '{
    "model": "openai.gpt-oss-120b",
    "input": [
        {"role": "user", "content": "Hello! How can you help me today?"}
    ]
}'

Stream responses

To receive response events incrementally, choose the tab for your preferred method, and then follow the steps:

OpenAI SDK (Python)
# Stream response events incrementally using the OpenAI SDK
# Requires OPENAI_API_KEY and OPENAI_BASE_URL environment variables

from openai import OpenAI

client = OpenAI()

stream = client.responses.create(
    model="openai.gpt-oss-120b",
    input=[{"role": "user", "content": "Tell me a story"}],
    stream=True
)

for event in stream:
    print(event)
HTTP request

Make a POST request to /v1/responses with stream set to true:

# Stream response events incrementally
# Requires OPENAI_API_KEY and OPENAI_BASE_URL environment variables

curl -X POST $OPENAI_BASE_URL/responses \
   -H "Content-Type: application/json" \
   -H "Authorization: Bearer $OPENAI_API_KEY" \
   -d '{
    "model": "openai.gpt-oss-120b",
    "input": [
        {"role": "user", "content": "Tell me a story"}
    ],
    "stream": true
}'

Using the Responses API on the bedrock-runtime endpoint

The Responses API on bedrock-runtime uses the same request and response format as on bedrock-mantle, so the OpenAI SDK works against either one. What changes is the base URL, the model IDs, the permissions, and a small number of behaviors described in this section.

Base URL and paths

Set your base URL to https://bedrock-runtime.region.amazonaws.com/openai/v1. The API is served on the following paths:

  • POST /openai/v1/responses – create a response.

  • GET /openai/v1/responses/{id} – retrieve a stored response.

  • POST /openai/v1/responses/{id}/cancel – cancel a response that is still in progress.

  • DELETE /openai/v1/responses/{id} – delete a stored response.

Model IDs

Name a cross-Region inference profile as the model, not a foundation model ID. The OpenAI GPT models use the us. and global. profiles in the commercial Regions and the us-gov. profiles in the AWS GovCloud (US) Regions — for example, us.openai.gpt-5.6-sol. In-Region inference isn't available for these models on this endpoint. For the profile ID of each model, see its model card in Models at a glance, and for how routing works, see Route model inference requests across AWS Regions with cross-Region inference.

Permissions

Creating a response authorizes two resources: bedrock:InvokeModel (or bedrock:InvokeModelWithResponseStream) on the inference target, as any inference request does, and bedrock:InvokeModel on your account's default project. Retrieving, canceling, and deleting a stored response authorize bedrock:GetInvoke, bedrock:CancelInvoke, and bedrock:DeleteInvoke respectively, each on the project. Individual response IDs are not IAM resources.

Two condition keys let a policy on either resource constrain the other. The inference-target authorization carries bedrock:ProjectArn, and the project authorization carries bedrock:ModelArn, valued at the inference profile or foundation model your request named — never the destination models that a cross-Region profile routes to. For policy examples, see Prerequisites for running model inference.

Behavior differences

  • Requests are always synchronous. background=true is rejected with a 400 error. The store parameter is unaffected and keeps its default of true, so stored, multi-turn conversations work normally.

  • model is required on every request, including one that supplies previous_response_id. This differs from the OpenAI Responses API specification and from bedrock-mantle, where the model can be omitted and inherited from the previous response. The model is part of what the request is authorized against, so it has to be named in the request itself.

  • Server-side tool use and pre-configured tools aren't available, including web search. Client-side tool use works on both endpoints.

  • Only the default project is supported. The OpenAI-Project header is accepted only as default or as your own default project ARN; any other value is rejected. See Projects (OpenAI-compatible).

  • Application inference profiles aren't supported. A request that names one as its inference target is rejected with a 400 error. System, geographic, and global inference profiles work normally.

  • Guardrails don't apply to the Responses API. To apply a guardrail to a GPT model on this endpoint, call the Converse API instead.

  • A stored response belongs to the AWS Region that served it. Retrieving, canceling, or deleting it, and continuing the conversation with previous_response_id, are all handled by that Region. A response ID that can't be found — because it never existed, belongs to another account, or was never stored — returns the same 404 error in every case.

Monitoring and cost

Because every request is synchronous, CloudWatch metrics and model invocation logging work for the Responses API the same way they do for the other inference APIs on this endpoint, including for streaming requests. Usage is attributed to the inference target, exactly as it is for Converse and InvokeModel — the default project is never the billing anchor. See Track usage and costs in Amazon Bedrock.