AWS introduces OpenAI GPT-5.6 in-country routing on Amazon Bedrock in India
The integration provides localized cross-Region inference between Mumbai and Hyderabad, complying with local data residency laws and offering prompt caching with zero data retention.
Original cover art, generated for this story. THE VISSION does not republish third-party press imagery.
- AWS launched in-country geographic cross-Region inference for OpenAI's GPT-5.6 models on Amazon Bedrock in India, according to an official announcement.
- The system routes inference requests between AWS Mumbai and Hyderabad datacenters, ensuring all data processing remains within India to meet compliance standards.
- The service features prompt caching with up to a 90% discount on cached input tokens, reasoning depth control, and a Zero Data Retention model.
Amazon Web Services (AWS) announced the availability of OpenAI’s flagship GPT-5.6 models — Terra and Luna — on Amazon Bedrock for in-country geographic inferencing in India, according to a technical post on the AWS Machine Learning blog. The new integration addresses strict local data residency requirements by keeping all inference requests and data within India's borders.
To ensure compliance with local regulations, Bedrock uses India geographic cross-Region inference, routing requests dynamically between the Mumbai (ap-south-1) and Hyderabad (ap-south-2) regions. This setup allows enterprise developers in highly regulated sectors like financial services, healthcare, and the public sector to deploy GPT-5.6 models at scale without risk of data exporting.
Both GPT-5.6 Terra and Luna support a 1-million-token context window, multi-modal image inputs, and text outputs. The service also includes developer-focused efficiency features such as prompt caching — which offers up to a 90% cost discount on cached input tokens for repeated context above 1,024 tokens — and reasoning depth control. To further protect sensitive corporate data, the deployment operates under a strict Zero Data Retention (ZDR) model, preventing model inputs and outputs from being stored by default.
For multinational cloud providers and AI developers in India, compliance is a massive hurdle. AWS offering OpenAI's GPT-5.6 models natively on Bedrock with strict in-country routing between Mumbai and Hyderabad removes a major blocker for enterprise adoption, allowing Indian financial and healthcare enterprises to deploy frontier-tier reasoning without violating domestic data residency laws.