Amazon Web Services (AWS) has expanded access to OpenAI’s GPT-6 models on its Amazon Bedrock platform across five Asia Pacific regions. This includes models GPT-6 Sol and GPT-6 Luna, which are now reachable through global cross-Region inference from Thailand, Malaysia, Singapore, Indonesia, and Taiwan. GPT-6 Astra is available exclusively from Singapore. This announcement was detailed in a blog post on the AWS Machine Learning Blog on October 12, 2026.
The blog post provides a configuration guide for users in the mentioned countries, explaining how to initiate API calls, set permissions, and monitor inference from the corresponding source regions. AWS noted that this cross-regional access allows broader capacity usage compared to single-region inference, which can become limited during high-demand periods.
Global Cross-Region Inference Profiles
According to AWS, users can access the models via specific profile IDs on the bedrock-runtime endpoint. The profiles include global.openai.gpt-6-sol and global.openai.gpt-6-luna, with processing possible from the regions in Singapore, Jakarta, Malaysia, Thailand, and Taipei. The model global.openai.gpt-6-astra is accessible only from Singapore. These profiles facilitate processing in various commercial AWS Regions where the models are deployed, listed under system-defined inference profiles in the Amazon Bedrock console.
With global cross-Region inference, users submit a profile ID rather than directly referencing the model ID. AWS manages the processing in the appropriate destination region while keeping billing and monitoring tied to the source region.
AWS described the three models by their intended use: GPT-6 Sol focuses on complex reasoning and coding tasks, GPT-6 Luna handles repetitive tasks at scale, and GPT-6 Astra is designed for intricate reasoning, coding, and long-document workflows. All models support text and image inputs and offer advanced prompt capabilities.
Access Paths and Configuration Requirements
Each model can be accessed through three paths on the bedrock-runtime: the Amazon Bedrock Converse API, the OpenAI Responses API, and the OpenAI Chat Completions API. Access requires an IAM identity with the necessary permissions and an activated source region; while Singapore is activated by default, the other mentioned regions require user activation. The blog post also provides guidelines for teams migrating from existing OpenAI clients, emphasizing the need for specific parameter configurations.
Monitoring and quota tracking occur in the source region, and usage is calculated in tokens per minute. Each input token counts towards the quota, while output tokens count at a higher rate. AWS recommends that organizations requesting increased quota do so in advance of critical production efforts.
Retention policies for automated offline abuse detection will apply in the destination regions for cross-Region inference requests, with customers able to request a zero data retention option through their AWS account management teams. The blog post advises those with data residency requirements to consult internal compliance teams early in the implementation process.
Going forward, AWS emphasizes the importance of configuring models correctly to avoid unexpected errors or limitations when using cross-Region capabilities.
