Integrating Claude Opus 5.5 with Amazon Bedrock

One-sentence takeaway

AWS documentation lists the model IDs, regional routing options, and API support for Claude Opus 5.5 on Bedrock Runtime and Bedrock Mantle. It also provides an example request to the Runtime Messages API using the Anthropic Python SDK.

Use cases

• Suitable for: Configuring Claude Opus 5.5 on Amazon Bedrock; choosing an inference route based on data residency requirements; confirming which endpoint and API to use.

• Not suitable for: Inferring Anthropic API configuration outside AWS, model performance, or specific prices. The documentation provides a complete example only for the Python Messages API.

• Applicable model version: Claude Opus 5.5; the model ID in the documentation is anthropic.claude-opus-5-5.

• Applicable clients, agents, or APIs: Amazon Bedrock bedrock-runtime and bedrock-mantle endpoints. Runtime supports Messages, Converse, and Invoke; Mantle supports Messages. Neither supports Responses or Chat Completions, and Mantle does not support Converse or Invoke. This API comparison is based on the explicit support icons in the page's table.

• Recommended reasoning level and parameters: The model page says adaptive thinking is always enabled and cannot be turned off. Available effort levels are low, medium, high, xhigh, and max; the default is medium. The integration example below does not set a reasoning level.

Ready-to-use content

The code below is based on the Runtime Messages API example on the AWS page. First, prepare an AWS account and a long-term Amazon Bedrock API key, then install the dependencies specified on the page:

pip install -U anthropic aws-bedrock-token-generator

Save the example as bedrock-first-request.py, then run it:

from anthropic import Anthropic
from aws_bedrock_token_generator import provide_token

token = provide_token(region="us-east-1")

client = Anthropic(
    base_url="https://bedrock-runtime.us-east-1.amazonaws.com/anthropic",
    api_key=token,
)

response = client.messages.create(
    model="global.anthropic.claude-opus-5-5",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Can you explain the features of Amazon Bedrock?"}],
)

print(response)

The page's setup steps separately list this environment variable: AWS_BEARER_TOKEN_BEDROCK="<provide your Bedrock API key>". The code example itself generates a token with provide_token(region="us-east-1") and passes it to the SDK's api_key. Configure credentials according to the authentication flow you choose; do not put a real key in code or commit it to version control.

Test or workflow steps

1. Prepare an AWS account; the page says new users must first sign up for an AWS account.

2. Generate a long-term API key in the Amazon Bedrock console.

3. Confirm that Python is installed, then install anthropic and aws-bedrock-token-generator.

4. Use a regional endpoint for bedrock-runtime. The example uses us-east-1; the Runtime URL format is https://bedrock-runtime.{region}.amazonaws.com, and the Anthropic SDK example's base_url ends in /anthropic.

5. For the Messages API, use global.anthropic.claude-opus-5-5 as the example model ID and send a request.

6. If using another API or endpoint, check the model ID and API support table below first; do not treat the Runtime example URL as a Mantle URL.

Original evidence and data

• Model release date: 2026-09-22; lifecycle status: Active; context window: 1M tokens; maximum output: 128K tokens; knowledge cutoff: June 2026.

• Runtime model ID: anthropic.claude-opus-5-5. Regional inference IDs: us.anthropic.claude-opus-5-5, eu.anthropic.claude-opus-5-5, au.anthropic.claude-opus-5-5, and jp.anthropic.claude-opus-5-5; global inference ID: global.anthropic.claude-opus-5-5. The table lists the Runtime In-Region endpoint URL as N/A.

• Mantle model ID: anthropic.claude-opus-5-5; URL template: https://bedrock-mantle.{region}.api.aws/anthropic/v1/messages. The page's example region, us-east-1, maps to https://bedrock-mantle.us-east-1.api.aws/anthropic/v1/messages.

• Data residency notes: the us. route keeps data within the United States and Canada; eu. limits it to the EU; au. to Australia; jp. to Japan; and global. allows global routing without residency constraints.

• The page recommends using bedrock-runtime for new applications whenever possible. Service tiers are Standard (pay per token), Priority (faster responses at a price premium), Flex (lower cost, suited to non-time-sensitive tasks), and Reserved (account-level reserved throughput, which requires contacting the AWS team to enable).

• The page identifies this model as a third-party model offered and billed through AWS Marketplace; charges appear on the AWS bill and in Cost Explorer under the model provider's name. For specific prices, see the AWS Bedrock Pricing page; the model card does not list price figures.

Limitations

• Regional availability varies by endpoint. The documentation also lists three inference options: In-Region, Geo Cross-Region, and Global Cross-Region. Before deployment, check the current AWS regional availability and quota pages for the target region; the model ID alone does not establish that the model can be called in a particular region.

• Runtime and Mantle support different APIs: the model card table shows that Runtime supports Messages, Converse, and Invoke, while Mantle supports only Messages; both mark Responses and Chat Completions as unsupported. The SDK example demonstrates only the Runtime Messages API.

• The documentation page provides no comparable price figures, request latency, or benchmark results. It cannot establish the specific magnitude of cost differences or task performance.

• Feature support listed in the model capabilities table describes Bedrock platform capabilities; it does not mean every region, account, or request configuration has those features enabled. Permissions, quotas, and region availability must still be checked separately.

Reproduction notes

Follow the official steps to install the dependencies, prepare a Bedrock API key in the AWS console, and choose a region where the model is available. Use the Python example on the page to call the us-east-1 Runtime Messages API, and save the actual response along with the SDK version, Python version, and credential method used. When switching to Mantle or a regional inference ID, update the configuration to match the corresponding endpoint template and residency requirements, and verify regional availability before making the call.