Amazon Bedrock now supports OpenAI’s GPT-5.6 Terra and Luna models with inference data kept inside India. AWS routes requests between its Mumbai and Hyderabad regions, allowing customers to draw on capacity in both locations without sending prompts or outputs outside the country.
Both models accept text and images, return text, and provide a one-million-token context window. Customers can call them through OpenAI-compatible Responses and Chat Completions interfaces or Amazon Bedrock’s Converse API. The geographic cross-region profile automatically chooses capacity within India when demand changes.
The release is aimed at organizations with local processing requirements, including financial services, healthcare, and the public sector. Cross-region inference is mainly a capacity and availability mechanism: it does not mean every workload automatically satisfies a company’s broader legal, security, or compliance obligations.
Developers must use one of the supported India regions as the source and select the relevant inference profile. Model and regional availability can change, so AWS directs customers to its current availability documentation. The practical change is that Indian teams can now use these OpenAI models through Bedrock while maintaining in-country processing at the inference layer.