AWS Puts OpenAI’s GPT-5.6 Terra and Luna in GovCloud, but Not Its Flagship
The move gives regulated U.S. workloads two lower-cost OpenAI model tiers inside Bedrock’s government regions, while preserving a meaningful capability boundary around GPT-5.6 Sol.
Listen to this story
The audio brief
Story brief
3 key pointsThe GovCloud rollout gives U.S. government customers access to OpenAI’s Terra and Luna through AWS, but with a meaningful capability ceiling: flagship Sol is unavailable. Both models offer million-token context windows and 90%-discounted prompt caching, while Bedrock’s compatibility with the OpenAI Responses API may ease migration for existing applications. Access requires accepting OpenAI’s license through a linked...
- 01
Terra targets balanced production workloads; Luna is optimized for faster, high-volume inference.
- 02
The models are available in GovCloud US-East and US-West via Bedrock’s bedrock-mantle endpoint.
- 03
Both support up to 1 million tokens of context and prompt caching with a stated 90% discount on reused context.
AWS has made OpenAI’s GPT-5.6 Terra and Luna generally available through Amazon Bedrock in its GovCloud US-East and US-West regions, opening a route for regulated U.S. government workloads to use the two models without leaving AWS’s government-cloud environment. The release does not include GPT-5.6 Sol, OpenAI’s flagship tier.
Two models enter the boundary
The choice of models defines the launch. OpenAI positions Terra as a balanced production model, while Luna is its faster option for high-volume inference. Sol remains the tier intended for the hardest coding, cybersecurity and scientific workloads in supported commercial Bedrock regions.
GovCloud is AWS’s isolated cloud environment for U.S. government agencies and organizations handling regulated workloads. AWS says its regions are physically and logically administered by U.S.-citizen personnel and are designed to help customers address requirements associated with FedRAMP High, Defense Department Impact Levels 4 and 5, criminal-justice data and U.S. export controls.
A long-context option with cost controls
Both models support context windows of up to 1 million tokens. That capacity can place large codebases, long documents or extended agent histories in one request, rather than requiring an application to divide that material into smaller prompts.
AWS also says Terra and Luna support prompt caching with explicit cache breakpoints. Reused context receives a 90% discount, according to AWS, which is relevant to systems that repeatedly send the same instructions, records or reference material with each request.
The integration path
The models are available through Bedrock’s bedrock-mantle endpoint and the OpenAI Responses API. AWS assigns the identifiers openai.gpt-5.6-terra and openai.gpt-5.6-luna, and developers can use the OpenAI software development kit while directing its base URL to the Bedrock endpoint.
That setup can reduce the work of moving an existing OpenAI application onto AWS infrastructure. AWS documents server-side tool calling, projects and prompt caching among Terra’s supported Bedrock features, extending the use case beyond a basic text-generation endpoint.
A separate agreement step
Access is not enabled solely from a GovCloud account. Customers must first accept the model’s end-user license agreement through a standard AWS account linked to their GovCloud account, then enable the model from Bedrock’s Model Access page inside GovCloud.
For data handling, AWS says prompts and completions sent to GPT-5.6 through Bedrock are not used to train models and are not shared with OpenAI. The company also says processing can remain inside either of the named GovCloud regions.
A month after the commercial release
The GovCloud release follows commercial Bedrock availability for the full GPT-5.6 family by one month. It gives AWS a way to offer OpenAI models through the same Bedrock service used by its government customers, while the omitted flagship leaves a clear limit on the capability set now available in GovCloud.
Editorial analysis
Our Read
Our read: This is chiefly a distribution and operating-model decision, not a full OpenAI rollout for government workloads. AWS is offering the two GPT-5.6 tiers aimed at repeatable production and high-volume work, plus a large context window and discounted reused context. But Sol’s absence means agencies cannot treat GovCloud availability as access to the entire family. The next useful signal will be whether AWS adds Sol or whether Terra and Luna become the durable default for document-heavy and tool-using government systems.
