AWS adds Bedrock runtime APIs and cross-region inference for OpenAI in GovCloud
AWS announced on October 30 that Amazon Bedrock supports OpenAI GPT-5.6 Terra and Luna on its bedrock-runtime endpoint in AWS GovCloud (US), with Responses, Chat Completions and Converse API support. The announcement also adds Cross-Region Inference within the GovCloud regions.
This is an expansion of endpoint and API support, not the first availability of those models in GovCloud. AWS announced Terra and Luna availability in GovCloud US-West and US-East on August 24, using its next-generation inference engine and the bedrock-mantle endpoint.
AWS says Cross-Region Inference routes requests across both GovCloud regions to provide more throughput and redundancy without customers managing capacity separately across those regions. That routing remains within the GovCloud regions described in the announcement; it should not be read as a guarantee about every service or data flow in a customer’s system.
With the native OpenAI APIs running on bedrock-runtime, AWS says usage can appear in Bedrock model invocation logging, deliverable to Amazon S3 or Amazon CloudWatch Logs. CloudWatch metrics cover invocation counts, token counts, latency, throttles and errors. These are supported monitoring features, not a claim that every customer has already enabled every log.
AWS also says usage is itemized in Cost Explorer and the AWS Cost and Usage Report so customers can attribute spending by model. The announcement does not publish a token price, promise a discount or establish that any specific procurement agreement covers a customer’s intended workload.
Public-sector teams should distinguish service availability from authorization for a particular deployment. The verified change is the new API and routing support described by AWS. It does not by itself establish that an agency’s application, selected configuration, workstation or entire data pipeline meets every security or regulatory requirement.