Reduce AI agent costs on Bedrock with new prompt caching
In short: Amazon Bedrock now supports explicit and implicit prompt caching for OpenAI models, allowing you to store and reuse static prompt prefixes like tools, instructions, and conversation history. By avoiding the re-processing of identical tokens across turns, you can significantly lower input costs and improve responsiveness in complex agent workflows.
This summary was generated automatically by AI from AWS Machine Learning's publication. It is our own text, not a copy of the original — facts, figures and quotes belong to the source, linked above and below.
What changed?
- 1Added explicit prompt_cache_breakpoint for defining static content boundaries.
- 2Introduced implicit caching for automatic management of growing conversation history.
- 3Support for additional_tools to inject new functions into existing sessions without invalidating the cache.
- 4Compatible with GPT-6 Astra, GPT-6.1 Sol, and GPT-5.6 Luna on Bedrock.
| Parameter | Before | Now |
|---|---|---|
| Input pricing | Not specified | Reduced via cache reuse (up to 62% in tests) |
| Supported Models | Not specified | GPT-6 Astra, GPT-6.1 Sol, GPT-5.6 Luna |
Why it matters
This is critical for companies running complex AI agents where tool definitions and system instructions are large. By caching these, you stop paying to re-read them every time an agent makes a turn or executes a function.
What it means for AI agents and contact centers
When building voice agents that require high responsiveness, implementing these caching patterns can significantly reduce latency in multi-turn conversations. Compare the cost-efficiency and latency gains of explicit versus implicit caching for your specific agent flow, especially when using models like GPT-6 Astra.
🧪 Worth Testing
It offers a direct path to lowering operational costs and improving agent latency in high-volume voice interactions.
Sources
- AWS Machine LearningOfficialPrimary sourceOriginal article →„Groundhog Day for AI agents: Prompt caching patterns for using OpenAI models on Amazon Bedrock“26 Oct 2026, 19:27
- Published by source
- 26 Oct 2026, 19:27
- Found by our system
- 26 Oct 2026, 19:51
- Summary generated
- 26 Oct 2026, 19:51
This article was written by AI from the original source. Facts, numbers and prices come from the source; missing values are marked “Not specified”. Legal notice, copyright and privacy