Amazon Bedrock has expanded its support for Anthropic’s Claude models to include in-country inferencing in India. This means users in India can now process data locally using models like Claude Opus 5, Claude Sonnet 5, and Claude Haiku 4.5.
The India regional endpoint is served through geographic cross-Region inference, which routes requests only between the Mumbai and Hyderabad Regions. This setup helps keep data within India and complies with local data rules.
Cross-Region inference profiles let users distribute inference across multiple AWS Regions without managing capacity in each one. Requests originate in the source Region and are routed to the destination Regions defined in the profile. This improves performance during high traffic.
Data sent for inference remains in the source Region, and Amazon Bedrock uses a zero data retention model. Customer data is not stored in the destination Region, and all data in transit is encrypted. Billing and monitoring are also tracked in the source Region.
Users can access Claude models through the Amazon Bedrock console or programmatically via APIs. The console offers a no-code way to test prompts and adjust parameters. APIs like the Anthropic Messages API and Amazon Bedrock InvokeModel API support the India geographic profile.
To get started, users need an active AWS account with Bedrock access, the Anthropic SDK installed, and proper permissions. Practical examples include using Python SDKs to invoke models and monitor usage via CloudWatch.
This update helps build scalable, resilient AI applications within India, supporting local data laws and improving response times for users in the country.