Introducing explicit prompt caching for OpenAI GPT-5.6 models on Amazon Bedrock

This article introduces explicit prompt caching for OpenAI GPT-5.6 models on Amazon Bedrock, a feature that provides precise control over which prompt portions are cached and reused across requests.

• GPT-5.6 Sol, Terra, and Luna now generally available on Amazon Bedrock with pay-per-token pricing and AWS security controls
• Explicit prompt caching enables 90% discount on cached input tokens, with cache writes billed at 1.25x standard rate
• Cached prefixes remain available for 30 minutes, ideal for agentic workflows with repeated system instructions and tool definitions
• Two caching modes: implicit (automatic breakpoints) and explicit (user-controlled boundaries) for different workload patterns
• Migration path from GPT-5.5/5.4 requires model ID change and adoption of explicit cache breakpoints with consistent cache keys
• Responses API provides OpenAI SDK compatibility with typed streaming events and server-managed conversation state
• Monitoring via usage.input_tokens_details fields (cached_tokens and cache_write_tokens) confirms write-once, read-many patterns

Explicit prompt caching on GPT-5.6 optimizes costs for applications that repeatedly send identical instructions and references, requiring minimal code changes for existing Bedrock users.