OpenAI GPT-5.6 Terra and Luna now available on Amazon Bedrock in AWS GovCloud (US)

Amazon Web Services announced the general availability of OpenAI’s GPT-5.6 Terra and Luna models on Amazon Bedrock within AWS GovCloud (US-West) and AWS GovCloud (US-East). These models are designed to enhance performance and cost efficiency, with Terra offering GPT-5.5-level capabilities at half the cost and Luna providing faster, more affordable inference. The release targets workloads requiring high reliability and security, such as autonomous coding agents and advanced cybersecurity research.

GPT-5.6 introduces a one-million-token context window on Amazon Bedrock, enabling users to process entire codebases, lengthy documents, or multi-turn agent interactions in a single request. This feature reduces the need for manual chunking and minimizes information loss, improving response accuracy and coherence. The expanded context window supports complex workflows, including genomics and biology analyses, where large datasets are common.

The models include prompt caching with explicit cache breakpoints, allowing repeated context in agentic workflows to be billed at a 90% discount. This cost-saving measure prevents compounding expenses as usage scales, making the models more economical for long-running or iterative tasks. The discount applies to cached prompts, reducing operational costs for repetitive or high-volume inference requests.

Users can access GPT-5.6 Terra and Luna through the Amazon Bedrock Console or the Responses API on the bedrock-mantle endpoint. For regional availability details, AWS directs users to its Bedrock regional availability page. Additional documentation and a launch blog post provide further guidance on implementation and use cases.