Amazon Bedrock now supports in-region inference for Claude Opus 5 and Claude Sonnet 5 in Seoul, according to AWS. Claude Sonnet 5 is also available for in-region inference in Singapore. The AWS Region identifiers are ap-northeast-2 for Seoul and ap-southeast-1 for Singapore. The distinction matters when selecting a model: AWS lists both models for Seoul, while its Singapore announcement covers Sonnet 5.

What the regional boundary covers

For in-region inference, Amazon Bedrock processes a request within the single Region that is called. The input prompt and output result remain in that Region throughout the request. For a team choosing where an inference request is processed, the called Region is therefore as relevant as the model name. The announcement describes the location of request processing and of the prompt and result during that request.

AWS says this mode has no routing layer. Throughput is bounded by the capacity of the called Region. That gives the regional choice an operational consequence alongside its data-processing consequence: capacity for a request is tied to the Region receiving it. The announcement does not provide a benchmark comparison between Opus 5 and Sonnet 5; its concrete guidance concerns availability, regional processing, interfaces and regional operating limits.

How the models are reached

The in-region models are available on the bedrock-runtime endpoint through direct model IDs such as anthropic.claude-opus-5 and anthropic.claude-sonnet-5. For these models, that endpoint supports Anthropic’s Messages API as well as Amazon Bedrock’s InvokeModel and Converse APIs. The model ID identifies the model, while the Region called determines which of the announced regional options is being used.

AWS also says Claude models can be accessed in the Amazon Bedrock console’s text playground without writing code or setting up an SDK. This gives readers a console route to the models alongside the named runtime APIs. Whether using the playground or an API, the availability stated in this announcement remains specific: Opus 5 and Sonnet 5 in Seoul, and Sonnet 5 in Singapore.

Quotas, pricing and records follow the called Region

Requests are subject to per-Region service quotas and the standard on-demand pricing of the Region called. Quota consumption, CloudWatch metrics and CloudTrail log entries are scoped to that Region too. Anyone checking usage or operational records for these requests should therefore use the called Region as the frame of reference. The same regional choice governs where the request is processed, which Region’s capacity constrains throughput, and where the associated quotas and records are counted.

For readers assessing local data-processing requirements, the precise AWS claim is that the input prompt and output result stay in the single called Region for the duration of the request. Its announcement of the Seoul and Singapore availability sets out that boundary together with the supported models and interfaces.