Amazon Bedrock adds structured outputs for Claude models
Amazon Bedrock has introduced outputs based on JSON Schema and strict validation of tool parameters on the bedrock-runtime and bedrock-mantle endpoints. Support covers all Claude models available on the respective endpoint.
Amazon Bedrock now supports structured outputs for all Claude models available through the bedrock-runtime and bedrock-mantle endpoints. Both allow users to configure the response format using JSON Schema and enable strict validation of tool parameters. According to AWS, the structure is enforced during generation through constrained decoding, which restricts the choice of subsequent tokens according to the schema.
The feature is intended for tasks such as extracting data from documents, populating product catalogs, or routing customer support requests. The strict: true parameter enables strict adherence to the tool input schema. Support also includes batch inference; on the bedrock-runtime endpoint, it also works with cross-Region inference. According to AWS, limiting the response to the required fields may reduce output token usage and costs at high request volumes.
The first request with a new schema may take longer because the grammar needs to be compiled. The result is cached for the account for 24 hours from the first access; changing the schema structure triggers a new compilation. Enforcement also has limitations: the API accepts some conditions but does not check them during generation. According to AWS, these conditions must be validated in the application.
Why it matters
When passing model responses to other systems, the required structure can be set directly in the request. According to AWS, this reduces formatting corrections and repeated requests caused by schema violations. However, integration design needs to account for a slower first request and application-side validation of conditions that the service does not enforce.
Two audiences, two different impacts
What this means
For individuals
Developers using Claude models in Amazon Bedrock can directly define the response structure and the parameters of tool calls on both supported endpoints.
For a business
According to AWS, bulk processing of documents and catalogs may require fewer repeated requests and fewer output tokens. This may affect the operating costs of these processes; the source does not quantify the savings.
ProcessesCheck the original
Event sources
clearly official source · 1 publisher, 0 independent. We count feeds from the same owner only once.