Amazon Bedrock now offers global cross‑Region inference for OpenAI GPT‑6 Sol, Luna, and Astra models, allowing teams in Thailand, Malaysia, Singapore, Indonesia, and Taiwan to access the newest models when local capacity is insufficient. The service uses inference profiles that replace model IDs in API calls, routing requests to a destination region where the model is deployed. Users can invoke these profiles via the Converse API, OpenAI Responses API, or OpenAI Chat Completions API, passing the profile ID as the model. The article details how to configure calls, set IAM policies to grant bedrock:InvokeModel (and optionally bedrock:InvokeModelWithResponseStream for streaming), and use service control policies to allow inference while maintaining region controls. Quota is measured in tokens per minute in the source region, with output tokens consuming ten times the quota. The article also notes that data retention for abuse detection occurs in the destination region and that compliance considerations should be addressed early. This expansion improves model availability and resilience for Southeast Asian teams, reducing throttling and expanding access to advanced AI capabilities.
Accessing OpenAI GPT models on Amazon Bedrock with global cross-Region inference in Thailand, Malaysia, Singapore, Indonesia, and Taiwan
Amazon Bedrock now supports global cross‑Region inference for OpenAI GPT‑6 Sol, Luna, and Astra models, allowing teams in Thailand, Malaysia, Singapore, Indonesia, and Taiwan to access the newest models when local capacity is exceeded. The service lets users invoke these models from any of the five AWS regions, with the request routed to a destination region where the model is deployed. The article explains how to configure calls, set IAM and SCP permissions, and monitor usage.
Why it matters
Providing cross‑Region inference expands AI model availability and resilience for organizations in Southeast Asia, reducing bottlenecks and improving access to advanced language models.
Topic dossiers OpenAI