Next upSF Pitch Night by the AI Collective - #SFTechWeek
News

AWS adds in-region Claude 5 inference in Seoul and Singapore

Amazon Bedrock now supports in-region inference for Claude Opus 5 and Sonnet 5 in Seoul, and for Sonnet 5 in Singapore. The new options keep processing within one selected Region, according to AWS.

D
Sep 30, 2026 · 1 min read

Amazon Web Services added in-region Amazon Bedrock inference for Anthropic’s Claude Opus 5 and Claude Sonnet 5 in its Asia Pacific (Seoul) Region, and for Claude Sonnet 5 in Asia Pacific (Singapore). For those model-region combinations, AWS says customers can keep processing within one selected Region. The service uses the ap-northeast-2 Region code for Seoul and ap-southeast-1 for Singapore.

AWS says each in-region request is handled entirely within the customer-selected Region, with no cross-Region routing layer. According to the company, prompts and outputs stay there throughout the request lifecycle. That setup differs from the cross-Region inference profiles used in AWS’s recent Grok 4.7 addition to Amazon Bedrock.

That regional boundary also shapes capacity and service administration. AWS says throughput depends on capacity in the selected Region, and requests count against per-Region service quotas. Billing follows standard on-demand pricing for the Region called. Quota use, Amazon CloudWatch metrics and AWS CloudTrail logs are also scoped to that Region. AWS did not publish numerical capacity or default quota levels for the new model-region combinations.

Customers can call the models through the bedrock-runtime endpoint using the direct model IDs anthropic.claude-opus-5 and anthropic.claude-sonnet-5. AWS says they support the Anthropic Messages API along with Amazon Bedrock’s InvokeModel and Converse APIs.

AWS’s regional availability matrix lists Claude Opus 5 for in-region inference in Seoul but not Singapore, while Claude Sonnet 5 is listed in both Regions. Anthropic’s Bedrock documentation, observed on the launch date, still listed Seoul and Singapore as Global endpoint Regions and did not yet show the new regional endpoint availability. The supplied research could not determine whether the difference was due to documentation lag.

More news