GPT-5.6 on Bedrock adds an India-residency path, but only for Terra and Luna

AWS says Amazon Bedrock now supports GPT-5.6 Terra and GPT-5.6 Luna through India geographic cross-Region inference. The new profile identifiers are `in.openai.gpt-5.6-terra` and `in.openai.gpt-5.6-luna`, and AWS positions them for customers with India-local processing requirements. This is an availability and routing change for two existing GPT-5.6 tiers, not evidence of a new model release.

The geographic profiles can route a request between AWS's Mumbai and Hyderabad Regions to draw on capacity while keeping processing inside India. AWS says prompts and results may move between those two Regions, so this is regional residency rather than a guarantee that each request remains in its originating Region. AWS also states that its zero-data-retention default has an exception: content flagged by automated abuse-detection classifiers can be retained for offline abuse detection.

The scope is narrower than the full GPT-5.6 family. AWS lists Terra and Luna for the India-prefixed profiles, while its global profiles can include Sol, Terra, and Luna and may route requests to supported commercial Regions worldwide. OpenAI describes Terra as its balanced tier and Luna as its lower-cost tier, so teams that require India-only processing cannot simply substitute Sol without accepting the different routing model.

For developers already using OpenAI client libraries, AWS documents an OpenAI-compatible Responses API and Chat Completions API on the Bedrock runtime endpoint. The practical migration is therefore closer to changing the endpoint, credentials, and model parameter to an AWS inference-profile ID than rewriting an application around a new request format. That still warrants integration testing: AWS-specific authentication, quotas, observability, and profile selection become part of the deployment contract.

The update also sits alongside a separate distribution path for the full family. OpenAI says Sol, Terra, and Luna are now available in Kiro, where the models are used in development workflows built around requirements, designs, tasks, code review, and testing. Kiro availability does not remove the Bedrock residency distinction: it is a product integration, while the India profiles are an AWS routing and compliance option.

For technical teams, the actionable distinction is to choose the India profiles when locality is the controlling requirement and to evaluate global profiles only when their routing policy is acceptable. The sources verify the profile names, supported tiers, API compatibility, and stated retention caveat, but they do not establish workload-specific latency, throughput, pricing, or compliance sufficiency. Those remain deployment questions that need to be tested against the team's own AWS configuration and obligations.