Regional Cross-Region Inference Support Launched for OpenAI GPT-6 Models on Amazon Bedrock

Amazon Web Services has begun offering global cross-region inference support for OpenAI GPT-6 models on Amazon Bedrock for teams in Thailand, Malaysia, Singapore, Indonesia, and Taiwan in order to overcome local capacity constraints.

Global Cross-Region Inference Support

Inference operations performed in a single region can slow down when demand exceeds local capacity. Therefore, teams in Thailand, Malaysia, Singapore, Indonesia, and Taiwan needed access to the latest OpenAI models beyond these capacity limits.

Amazon Bedrock delivers the OpenAI GPT-6 Sol and GPT-6 Luna models via global cross-region inference from five AWS regions, while providing the GPT-6 Astra model from the Asia Pacific (Singapore) region.

GPT-6 Model Family and Features

The GPT-6 family consists of three distinct models, each accessible via global cross-region inference. These models support text and visual inputs, prompt caching, and client-side tool use.

While GPT-6 Sol offers strong reasoning capability for complex tasks, GPT-6 Luna facilitates high-volume repeatable work. GPT-6 Astra manages complex reasoning, coding, and long-document work.

Routing and Management Operations

Global cross-region inference profiles provide broad access to Amazon Bedrock capacity by routing requests to supported commercial AWS regions where the model is deployed.

All management processes such as billing, quota tracking, AWS CloudTrail, and Amazon CloudWatch remain in the source region where the request was made.

Steps to Get Started with the Models

Users can make API calls by using the global cross-region inference profile ID instead of the model ID. This process can be performed via the Converse API, OpenAI-compatible APIs, or coding clients.

Prerequisites required for getting started include an IAM identity with permission to invoke the profile, enabling the source region, and up-to-date AWS SDK tools.