Amazon Bedrock has expanded its support for OpenAI GPT-5.6 models by introducing cross-region inference across more than 25 AWS Regions. This feature allows users to route requests to different Regions based on real-time compute availability, enhancing throughput and maintaining consistent performance under load. Three GPT-5.6 variants—Sol, Terra, and Luna—are now available with cross-region inference capabilities, each tailored for different balance of capability and cost. Users can select between geographic and global inference profiles, depending on their data residency needs and workload requirements. The geographic inference profiles keep data processing within a predefined geography, while the global profiles allow requests to be routed across all supported AWS commercial Regions where the model is deployed. This update enables organizations to scale their AI workloads while maintaining flexibility in compute resource allocation. Source: awsml
Cross-region inference (CRIS) in Amazon Bedrock operates through inference profiles, which define a model and the Regions to which Amazon Bedrock can route a request. Users call the profile from a source Region, and Amazon Bedrock routes the request to a destination Region using compute in that Region. This mechanism improves scalability by allowing requests to draw on a broader pool of compute rather than being bound to one Region’s available capacity. Geographic inference profiles restrict requests to a predefined set of Regions within a specific geography, ensuring data residency compliance for workloads that require processing within a particular area. In contrast, global inference profiles offer the widest capacity pool and are ideal for workloads without geographic processing constraints. Billing and quota consumption are tracked against the user’s account regardless of the backend Region that handled the request, ensuring a unified spending and throughput view. Source: awsml
The GPT-5.6 family on Amazon Bedrock includes general-purpose and specialized cybersecurity variants. This post covers the three general-purpose variants that support cross-region inference: Sol, Terra, and Luna. All three accept text and image inputs and return text, have a 1 million token context window, and support reasoning mode, server-side tool calling, and prompt caching. Users can call these models using the OpenAI Responses API, the OpenAI Chat Completions API, and the Amazon Bedrock Converse API. Streaming is supported through the Responses and Chat Completions API (stream=True), and ConverseStream. Amazon Bedrock inference profiles are logical identifiers passed instead of a raw model ID, allowing users to route requests based on predefined geography or global availability. Source: awsml
