インドでAmazon Bedrock経由でAnthropic Claudeモデルにグローバルなクロスリージョン推論でアクセス可能に

The adoption and implementation of generative AI inference has increased with organizations building more operational workloads that use AI capabilities in production at scale. To help customers achieve the scale of their generative AI applications, Amazon Bedrock offers cross-Region inference (CRIS) profiles. CRIS is a powerful feature that organizations can use to seamlessly distribute inference processing across multiple AWS Regions. This capability helps you get higher throughput while you’re building at scale and helps keep your generative AI applications responsive and reliable even under heavy load.

We are excited to introduce Global cross-Region inference for Amazon Bedrock and bring Anthropic Claude models in India. Amazon Bedrock now offers Anthropic’s Claude Opus 4.6, Claude Sonnet 4.6, and Claude Haiku 4.5 through Amazon Bedrock Global cross-Region inference (CRIS) for customers operating in India. These frontier models deliver a massive 1-million token context window and advanced agentic capabilities, allowing your applications to process vast datasets and complex workflows with unprecedented speed and intelligence. With this launch, customers using ap-south-1 (Mumbai) and ap-south-2 (Hyderabad) can access Anthropic’s latest Claude models on Amazon Bedrock while benefiting from global inference capacity and highly available inference managed by Amazon Bedrock. With global CRIS, customers can scale inference workloads seamlessly, improve resiliency, and reduce operational complexity. In this post, you will discover how to use Amazon Bedrock’s Global cross-Region Inference for Claude models in India. We will guide you through the capabilities of each Claude model variant and how to get started with a code example to help you start building generative AI applications immediately.

Core functionality of Global cross-Region inference

Global cross-Region inference helps organizations manage unplanned traffic bursts by using compute resources across inference capacity across *commercial AWS Regions (Regions other than the AWS GovCloud (US) Regions and the China Regions)* globally. This section explores how the Global cross-Region inference feature works and the technical mechanisms that power its functionality.

Understanding inference profiles

Global cross-Region inference is offered through Inference profiles. Inference profiles operate on two key concepts:

• Source Region – The Region from which the API request is made

• Destination Region – A Region to which Amazon Bedrock can route the request for inference

To use Anthropic models, Amazon Bedrock offers out of the box Global Inference profiles. For example:

• Opus 4.6:

• Sonnet 4.6:

• Opus 4.5:

• Sonnet 4.5:

• Haiku 4.5:

For customers in India using BOM (ap-south-1) and HYD (ap-south-2), the respective source and destinations would be as follows:

• Source -> Destination

BOM (ap-south-1) -> AWS commercial Regions

• HYD (ap-south-2) -> AWS commercial Regions

For information about considerations for choosing between Geographic and Global cross-Region inference, see Choosing between Geographic and Global cross-Region inference on the Amazon Bedrock User Guide.

Implementing global cross-Region inference

As of today, you can implement global CRIS for the following models.

For example, to use Global cross-Region inference with Anthropic’s Claude Opus 4.5, you should complete the following key steps:

• Use the global inference profile ID – When making API calls to Amazon Bedrock, specify the global Anthropic’s Claude Opus 4.5 inference profile ID (global.anthropic.claude-opus-4-5-20251101-v1:0) instead of a Region-specific model ID. This works with InvokeModel, InvokeModelWithResponseStream, and Converse and ConverseStream APIs.

• Configure IAM permissions – To enable Global cross-Region inference for your users, you must apply a three-part AWS Identity and Access Management policy to the role. For more information, see Configuring IAM policy for global cross-Region inference.

Global cross-Region inference for India’s peak demand seasons

Indian enterprises face unique challenges during high-traffic periods. Diwali shopping surges, Dussehra ecommerce spikes, Eid celebrations, Christmas festivities, tax filing deadlines, cricket tournaments, and festival seasons when customer engagement peaks dramatically. Global cross-Region inference provides customers in India with the throughput elasticity needed to handle these demand surges without degradation.During such peak periods ecommerce platforms, fintech applications, and customer service chatbots experience 3-5x the normal traffic. With Global CRIS, your applications automatically access inference capacity across AWS Commercial Regions, helping with

• Uninterrupted service during festival shopping peaks – When millions of customers simultaneously interact with AI-powered product recommendations and customer support, for example, during Diwali, Eid, Christmas, and regional celebrations

• Seamless tax season processing – Handles the surge in document analysis, form processing, and compliance queries during the July-September tax filing periods

• Festival campaign scalability – Deploys AI-driven marketing campaigns during festivals without capacity planning overhead

• Business continuity during cultural and sporting events – Maintains performance during cricket tournaments, election periods, and other events that drive massive engagement spikes

By routing requests globally, customers in India gain access to a significantly larger capacity pool—transforming from Regional Tokens per Minute (TPM) limits to global-scale throughput. This means that your generative AI applications remain responsive and reliable when your business needs them most, without the operational complexity of manual multi-Region orchestration or the risk of customer-facing throttling errors.

Step by step guidance for getting started with inferencing with Global cross-Region inference on Amazon Bedrock:

There are two approaches to infer the Global cross-Region supported models. Let’s understand these approaches in detail.

Approach 1: Inferencing on global CRIS models on AWS Console with Amazon Bedrock playgrounds – Asia Pacific (Mumbai) Region

The Amazon Bedrock playgrounds provide a visual interface to experiment on different models by using different configuration parameters. You can use playgrounds to test and compare different models by experimenting with prompts before integrating them into your application.

To get started using playgrounds, complete the following steps:

• Log in to the AWS Console and select region as Asia Pacific (Mumbai) Region.

• Navigate to Amazon Bedrock console

• On the side navigation menu for Infer, select Inference profiles.

• Global inference profiles will start with the keyword global.

• The model ID to invoke models via global CRIS is obtained via the corresponding Inference profile ID column from the System-defined inference profiles table. At the time of writing this blog post, it supports Anthropic Claude Opus 4.6, Claude Sonnet 4.6, Claude Haiku 4.5, and other foundation models.

• Select the Name of the model for example, Global Anthropic Claude Opus 4.6 and navigate to the screen where you will be shown the model details.

• Review the Inference profile overview of the chosen model from previous step. To perform inference in Chat/Text playground, choose Open In Playground button in the console.

• You can start inferencing the model on Amazon Bedrock chat / text playground, by trying out some prompts

Approach 2: Inferencing global cross-Region inference models programmatically

To invoke the global models programmatically, one can use InvokeModel, ConverseAPI for real-time requests, and InvokeModelWithResponseStream and ConverseStreamAPI for streaming workloads. The full source code demonstrating these invocation APIs is available at the GitHub repository aws-samples/sample-amazon-bedrock-global-cris.

Invoke Anthropic Claude model with global cross-Region inference using Converse API

Let’s understand the implementation of global cross-Region inference on global.anthropic.claude-opus-4-6-v1 model using Converse API. We recommend the Converse API for conversational applications with a unified interface.

import boto3

# Initialize Bedrock client for India region (Mumbai)
bedrock = boto3.client("bedrock-runtime", region_name="ap-south-1")

# Global CRIS model ID for Claude Opus 4.6
MODEL_ID = "global.anthropic.claude-opus-4-6-v1"

try:
    print("Invoking Claude Opus 4.6 via Global CRIS...")

    # Use Converse API for simplified interaction
    response = bedrock.converse(
        messages=[
            {
                "role": "user",
                "content": [{"text": "Explain cloud computing in 2 sentences."}],
            }
        ],
        modelId=MODEL_ID,
    )

    # Extract and display response
    response_text = response["output"]["message"]["content"][0]["text"]
    print("Response:", response_text)

    # Display token usage information
    usage = response.get("usage", {})
    print("Tokens used:", usage)

    if usage:
        print(f"Input tokens: {usage.get('inputTokens', 'N/A')}")
        print(f"Output tokens: {usage.get('outputTokens', 'N/A')}")
        print(f"Total tokens: {usage.get('totalTokens', 'N/A')}")

except Exception as e:
    print(f"Error: {e}")
    print("Please check your AWS credentials and region configuration.")

Code samples to invoke the Anthropic Claude model with global cross-Region inference with different API types

The code samples for InvokeModel, InvokeModelWithResponseStream, and Converse and ConverseStream APIs for Global CRIS models can be referenced as follows:

Model name

Inference profile ID

Invocation API

GitHub sample code

Global Anthropic Claude Opus 4.6

global.anthropic.claude-opus-4-6-v1

Converse

Code

ConverseStream

Code

InvokeModel

Code, Advanced Usage

InvokeModelWithResponseStream

Code, Advanced Usage

Global Anthropic Claude Sonnet 4.6

global.anthropic.claude-sonnet-4-6

Converse