Claude models now available through India regional endpoint on Amazon Bedrock

MUMBAI: Claude is staying closer to home. Amazon Web Services (AWS) has expanded the availability of Anthropic’s Claude models on Amazon Bedrock to India, allowing customers to use geographic cross-Region inference while keeping AI processing within the country.

AWS said customers in India can now access Claude Opus 5, Claude Sonnet 5 and Claude Haiku 4.5 through an India regional endpoint. The service uses geographic cross-Region inference between AWS’s Mumbai and Hyderabad regions, allowing enterprises to process their data within India while accessing a broader pool of computing capacity.

The development could be particularly relevant for organisations that have requirements around data localisation or prefer to keep sensitive workloads within a specific geography.

Under the India geographic inference profile, requests originating in the Mumbai region can be routed to Hyderabad, and vice versa. The system automatically distributes inference requests between the two India regions rather than limiting applications to the capacity available in a single location.

AWS said this approach can help maintain throughput and performance during periods of high demand while keeping inference within India’s geographical boundaries.

The requests and resulting outputs can move between the Mumbai and Hyderabad regions over AWS’s secure network, with encryption for data in transit. AWS said customer data is not stored in the destination region when using cross-Region inference and remains exclusively within the source region.

Amazon Bedrock also operates under a zero data retention model by default, meaning model inputs and outputs are not stored. AWS noted, however, that certain models can require human review if content is flagged by automated safety classifiers.

The company’s monitoring and billing structure also remains tied to the source region. Usage and quota consumption are tracked against the customer’s account in the source region, while Amazon CloudWatch and AWS CloudTrail log activity there as well.

The India geographic cross-Region inference capability is available through the Bedrock runtime endpoint and supports Anthropic’s Messages API, along with Amazon Bedrock’s InvokeModel and Converse APIs. It can also be used with features such as Amazon Bedrock Guardrails and intelligent prompt routing.

For enterprises, the change means Claude models can be incorporated into applications without requiring them to manage inference capacity separately across India’s two supported regions.

AWS said customers can also test the Claude models through the Amazon Bedrock console without setting up code or software development kits. Users can select an India inference profile in the Bedrock text playground, enter a prompt and test the model before integrating it into their applications.

Developers can access Claude models programmatically using the AWS SDK, AWS Command Line Interface, Anthropic’s Messages API or Amazon Bedrock’s InvokeModel and Converse APIs.

The move expands the options available to Indian organisations looking to deploy advanced AI while keeping their data processing within the country. It also gives AWS a broader platform for serving Claude workloads locally, combining India’s growing demand for generative AI with the need for performance, security and data residency.

For businesses weighing AI deployment, the message is fairly simple: Claude can now do more of its heavy lifting without the data having to leave India.