OpenAI GPT-6 Models Now on Amazon Bedrock in Asia-Pacific

Disclosure: this post contains affiliate links. If you buy through them we may earn a commission at no extra cost to you. Learn more.

TL;DR: Amazon Bedrock now offers OpenAI's GPT-6 models (Sol, Luna, and Astra) through global cross-Region inference in Thailand, Malaysia, Singapore, Indonesia, and Taiwan. The service routes requests across multiple AWS Regions to provide broader capacity access while keeping billing, quota tracking, and monitoring in the source Region.

What's New

Amazon Web Services has announced availability of OpenAI's GPT-6 model family on Amazon Bedrock with global cross-Region inference support across five Asia-Pacific regions. The release includes three models: GPT-6 Sol, GPT-6 Luna, and GPT-6 Astra, all accessible from Thailand, Malaysia, Singapore, Indonesia, and Taiwan.

GPT-6 Sol is designed for complex tasks requiring strong reasoning and coding capabilities. GPT-6 Luna targets focused, repeatable tasks suitable for high-volume processing. GPT-6 Astra handles complex reasoning, coding, and long-document work. All three models accept text and image input, support client-side tool use, and include prompt caching capabilities.

The global cross-Region inference feature addresses a key limitation of single-Region inference: throttling when demand exceeds local capacity. By routing requests across supported commercial AWS Regions where models are deployed, the service provides broader access to Amazon Bedrock capacity. Customers invoke models by passing a global cross-Region inference profile ID instead of a standard model ID.

According to the announcement, billing, quota tracking, AWS CloudTrail, and Amazon CloudWatch all remain in the source Region where the API call originates, while Amazon Bedrock selects the destination Region for processing. The models are accessible through three APIs: the Amazon Bedrock Converse API, the OpenAI Responses API, and the OpenAI Chat Completions API.

All three GPT-6 models support different reasoning effort levels. GPT-6 Sol and GPT-6 Luna accept six levels: none, low, medium, high, xhigh, and max. GPT-6 Astra accepts the same levels except "none."

Pricing and Availability

GPT-6 Sol and GPT-6 Luna are available through global cross-Region inference from all five AWS Regions: Asia Pacific (Singapore), Asia Pacific (Jakarta), Asia Pacific (Malaysia), Asia Pacific (Thailand), and Asia Pacific (Taipei). GPT-6 Astra is available only from Asia Pacific (Singapore).

Pricing information is available on the Amazon Bedrock pricing page, though specific per-token rates were not disclosed in the announcement. The service is currently available, with inference profile availability and Region support subject to change as AWS adds new Regions and profiles.

Singapore (ap-southeast-1) is enabled by default, while Jakarta (ap-southeast-3), Malaysia (ap-southeast-5), Thailand (ap-southeast-7), and Taipei (ap-east-2) are opt-in Regions that customers must activate first.

How to Try It

Developers need three prerequisites to get started: an IAM identity with permission to invoke the profile, activation of their source Region, and a current AWS SDK or CLI. For OpenAI-compatible examples, users must also install the OpenAI SDK and Amazon Bedrock token generator using pip.

The Amazon Bedrock console lists all three global cross-Region inference profiles under system-defined inference profiles. Developers can invoke models through code examples provided for both the Converse API and OpenAI Chat Completions API, with the ability to switch between models by changing the profile ID.

Who It's For

This release targets development teams and enterprises in Southeast Asia and Taiwan that need access to the latest OpenAI models beyond single-Region capacity limits. Organizations requiring high-volume AI processing, complex reasoning tasks, or long-document analysis will benefit from the expanded capacity and regional availability.

The service particularly addresses needs of teams experiencing throttling issues with single-Region inference, enabling them to maintain consistent access during peak demand periods.

Alternatives

Organizations seeking AI model access have several options beyond Amazon Bedrock. Replit offers AI-powered development environments with integrated model access. Companies can also access OpenAI models directly through OpenAI's API, though without the multi-region routing capabilities that Amazon Bedrock provides. Google Cloud's Vertex AI and Microsoft Azure OpenAI Service offer similar managed AI model platforms with their own regional distribution strategies.