OpenAI GPT-5.6 on Bedrock now keeps data in India

This announcement extends Amazon Bedrock's OpenAI GPT-5.6 model support to India with a geographic data-residency boundary. Previously, cross-Region inference for these models could route requests globally; now, organizations with strict local processing requirements can use the India profiles to keep inference within the country, routing only between the Mumbai and Hyderabad Regions. The move is particularly relevant for regulated industries like financial services and healthcare, where data sovereignty is a compliance concern.

Technically, the India geographic profiles work through inference profiles, which Amazon Bedrock uses to define routing constraints. Calls to these profiles track billing and quota against the source Region, and CloudWatch and CloudTrail logs also stay in the source Region, simplifying monitoring. Notably, the models support a 1-million-token context window and multimodal input, enabling complex workloads like invoice extraction or long-document processing without leaving the country.

For developers, Amazon Bedrock supports both OpenAI-compatible APIs (Responses and Chat Completions) and the Bedrock-native Converse API. Prompt caching, available under the India profiles, offers a 90 percent discount on cached reads and keeps cached prefixes warm for at least 30 minutes, which could reduce costs for RAG and agent workloads that repeat context. This combination of data residency, API flexibility, and cost-saving features may encourage more enterprise adoption of these models within India, though the actual business impact will depend on how organizations weigh performance against compliance requirements.