NEWS · MODELS · #2236
Amazon Bedrock adds structured outputs for Anthropic Claude models
Amazon Bedrock now supports structured outputs (schema-compliant JSON) for all Anthropic Claude models on both the bedrock-runtime and bedrock-mantle endpoints, using constrained decoding so generated tokens conform to a provided JSON Schema. The feature includes grammar compilation and 24-hour caching of compiled grammars, works with batch and streaming inference and cross-Region runtime, and aims to eliminate parsing errors and post-processing for production integrations.
KEY POINTS
- Amazon Bedrock now supports structured outputs (schema-compliant JSON) for all Anthropic Claude models on both the bedrock-runtime and bedrock-mantle endpoints, using constrained decoding so generated tokens conform to a provided JSON Schema.
- The feature includes grammar compilation and 24-hour caching of compiled grammars, works with batch and streaming inference and cross-Region runtime, and aims to eliminate parsing errors and post-processing for production integrations.
- Schema‑enforced, deterministic outputs reduce parsing errors, retries, and integration complexity, making Claude on Bedrock more reliable for production data-extraction, agentic workflows, and large-scale inference.
WHY IT MATTERS
Schema‑enforced, deterministic outputs reduce parsing errors, retries, and integration complexity, making Claude on Bedrock more reliable for production data-extraction, agentic workflows, and large-scale inference.