Unlocking AI Potential: A Guide to Cross-Region Inference for Canadian Organizations

Transforming Operations with Generative AI on Amazon Bedrock

Canadian Cross-Region Inference: Your Gateway to Global AI Innovation

Setting Up Cross-Region Inference for Canadian Organizations

Getting Started with Cross-Region Inference

Efficient Quota Management for Canadian Workloads

Migrating from Older Models to Claude 4.5

Choosing Between US and Global Inference Profiles

Conclusion: Embrace AI While Ensuring Data Governance

Meet the Authors: Your Expert Team at AWS

Transforming Canadian Organizations with Cross-Region Inference

Generative AI has revolutionized the way organizations operate, unleashing unprecedented opportunities for transformation and enhancing customer experiences. We’re thrilled to announce that Canadian customers can now leverage advanced foundation models, including Anthropic’s Claude Sonnet 4.5 and Claude Haiku 4.5, via Amazon Bedrock’s Cross-Region Inference (CRIS). This innovative approach enables Canadian organizations to access the latest models, enhancing operational efficiency and boosting AI initiatives.

Canadian Cross-Region Inference: Your Gateway to Global AI Innovation

Amazon Bedrock’s Cross-Region Inference (CRIS) profiles empower organizations to distribute inference processing seamlessly across multiple AWS Regions. This significantly increases throughput while ensuring that generative AI applications remain responsive and reliable, even when faced with high demand.

CRIS Profile Types:

1. Geographic CRIS: Automatically selects the optimal commercial Region within a specified geography for processing inference requests.
2. Global CRIS: Directs inference requests to supported commercial Regions worldwide, optimizing resources for higher model throughput.

All operations are conducted over AWS’s secure network, ensuring end-to-end encryption for data in transit and at rest. When inference requests are submitted from the Canada (Central) region, CRIS intelligently routes them to the most suitable destination, whether it’s a US region or a global option.

Cross-Region Inference Configuration for Canada

With CRIS, Canadian organizations now have access to foundation models faster, including Claude Sonnet 4.5, which boasts enhanced reasoning capabilities. This leads to improved capacity and performance during high-demand periods.

Inference Profile Options:

Getting Started with CRIS from Canada

1. Configure AWS Identity and Access Management (IAM) Permissions

Ensure your IAM role has the requisite permissions to invoke Amazon Bedrock models using CRIS profiles. Here’s an example policy for US cross-Region inference:

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": [
                "bedrock:InvokeModel*"
            ],
            "Resource": [
                "arn:aws:bedrock:ca-central-1::inference-profile/us.anthropic.claude-sonnet-4-5-20250929-v1:0"
            ]
        },
        {
            "Effect": "Allow",
            "Action": [
                "bedrock:InvokeModel*"
            ],
            "Resource": [
                "arn:aws:bedrock:*::foundation-model/anthropic.claude-sonnet-4-5-20250929-v1:0"
            ],
            "Condition": {
                "StringLike": {
                    "bedrock:InferenceProfileArn": "arn:aws:bedrock:ca-central-1::inference-profile/us.anthropic.claude-sonnet-4-5-20250929-v1:0"
                }
            }
        }
    ]
}

2. Use Cross-Region Inference Profiles

Configure your application with the necessary inference profile ID:

• Claude Sonnet 4.5 (US Regions): us.anthropic.claude-sonnet-4-5-20250929-v1:0
• Claude Haiku 4.5 (Global): global.anthropic.claude-haiku-4-5-20251001-v1:0

3. Example Code

To illustrate using the Amazon Bedrock Converse API with a US CRIS inference profile from Canada:

import boto3

# Initialize Bedrock Runtime client
bedrock_runtime = boto3.client(
    service_name="bedrock-runtime",
    region_name="ca-central-1"  # Canada (Central) Region
)

# Define Inference Profile ID
inference_profile_id = "us.anthropic.claude-sonnet-4-5-20250929-v1:0"

# Prepare the conversation
response = bedrock_runtime.converse(
    modelId=inference_profile_id,
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "text": "What are the benefits of using Amazon Bedrock for Canadian organizations?"
                }
            ]
        }
    ],
    inferenceConfig={
        "maxTokens": 512,
        "temperature": 0.7
    }
)

# Print the response
print(f"Response: {response['output']['message']['content'][0]['text']}")

Quota Management for Canadian Workloads

Quota management for CRIS operates at the source Region level (ca-central-1). Requests for quota increases apply to all inference requests from Canada.

Understanding Quota Calculations

Bear in mind the burndown rate, particularly for models like Claude Sonnet 4.5, which has a 5x burn down rate for output tokens. This means that 1 output token consumes 5 tokens from your quota. The total token calculation per request is:

Input token count + Cache write input tokens + (Output token count x Burndown rate)

Requesting Quota Increases

To request quota increases for CRIS in Canada, navigate to the AWS Service Quotas console and submit your requests based on projected usage.

Migrating from Older Claude Models to Claude 4.5

Organizations currently utilizing older Claude models should consider transitioning to Claude 4.5 to harness the latest model capabilities. Here’s a recommended migration strategy:

• Benchmark Current Performance: Establish metrics for existing models.
• Validate Performance: Test Claude 4.5 with real workloads and fine-tune prompts.
• Implement Gradual Rollout: Transition progressively to minimize risks.
• Monitor and Adjust: Track performance and adjust quotas accordingly.

Choosing Between US and Global Inference Profiles

Organizations can opt between US and Global inference profiles based on their individual requirements. US ES is ideal for those with US data processing agreements and those requiring high throughput.

Conclusion

Cross-Region Inference for Amazon Bedrock is a game-changer for Canadian organizations looking to leverage AI while adhering to data governance standards. This robust approach ensures faster access to advanced models, automatic scaling during peak times, and strict compliance with data regulations.

By adopting CRIS, Canadian organizations can innovate at lightning speed, meeting global standards while reinforcing data governance protocols. Start by reviewing your governance requirements and configuring IAM permissions, then test out the inference profile that best suits your needs.

About the Authors

This post benefitted from insights by Daniel Duplessis, Dan MacKay, Melanie Li, Serge Malikov, Saurabh Trikande, and Sharadha Kandasubramanian—experts in generative AI and compliance within AWS, bringing valuable perspectives to the evolving landscape of AI technologies.

Harness the power of Generative AI and transform your Canadian organization with Amazon Bedrock’s Cross-Region Inference today!