In its cloud AI foundation service, Amazon Bedrock now offers a one-hour TTL for prompt caching on select Anthropic Claude models, expanding from the prior default window of five minutes. This extension allows cached conversational contexts to persist longer, benefiting multi-turn conversations with infrequent user interactions or complex agent orchestration tasks.
The feature is intended to reduce repetitive prompt processing and optimize costs during extended or batch AI interactions. It is generally available across commercial and AWS GovCloud regions where the supported Claude model variants are accessible, billed at differentiated rates according to caching duration.
Source & further reading
Amazon Bedrock now supports 1-hour duration for prompt caching ↗Source published 2026-01-26. This report was prepared with AI assistance from the linked primary-source material. Claims and limitations should be checked against the original announcement or research.