Which GPT-6 models are now accessible and where?

Amazon Bedrock now provides three OpenAI GPT-6 generation models: GPT-6 Sol, GPT-6 Luna, and GPT-6 Astra, accessible via cross-region inference from five AWS regions in Southeast Asia and Taiwan, according to the official AWS Machine Learning blog.

The GPT-6 Sol and GPT-6 Luna models are available from the Asia Pacific (Singapore), Asia Pacific (Jakarta), Asia Pacific (Malaysia), Asia Pacific (Thailand), and Asia Pacific (Taipei) regions. In contrast, the GPT-6 Astra model is only accessible from the Asia Pacific (Singapore) region.

Cross-region inference allows users to access advanced AI models without geographical constraints, while simplifying cost and log management.

What features and reasoning levels are supported?

The three models accept inputs combining text and images, as well as client-side tool use and prompt caching, a feature useful for optimizing performance and reducing costs, AWS notes. The reasoning levels for GPT-6 Sol and GPT-6 Luna include six options: none, low, medium, high, xhigh, and max. Conversely, GPT-6 Astra does not accept the none level, suggesting a more task-oriented approach for complex operations from the start.

How does cross-region inference work, and what are its technical implications?

Cross-region inference enables routing requests to commercially available AWS regions where the model is deployed, thereby expanding access to Amazon Bedrock capacity. This approach ensures that users in the Asia Pacific (Jakarta), Asia Pacific (Malaysia), Asia Pacific (Thailand), and Asia Pacific (Taipei) regions can benefit from the models, even if they are not deployed locally. Billing, quota tracking, and logs via AWS CloudTrail and Amazon CloudWatch remain tied to the source region of the request, simplifying management for organizations.

The regions Jakarta, Malaysia, Thailand, and Taipei must be enabled individually before use, while Singapore is enabled by default. Cross-inference profiles are listed in the Amazon Bedrock console and can be invoked via the Converse API, the OpenAI Responses API, or the OpenAI Chat Completions API using the bedrock-runtime endpoint.

What this changes here

For businesses and government agencies in Benin and West Africa, this expansion of cross-region inference could simplify access to advanced AI models without requiring costly local deployments. Organizations in neighboring countries like Nigeria, Ghana, or Côte d'Ivoire, which already use AWS infrastructure, could explore partnerships with local players in Southeast Asia to integrate these models into their solutions while benefiting from reduced latency due to the geographic proximity of AWS regions.

Public administrations could also consider pilot projects using these models for tasks requiring high-level reasoning, such as complex data analysis or decision-making process automation. However, reliance on specific regions like Singapore for certain models could raise data sovereignty challenges, an increasingly pressing issue in Africa.