Amazon Bedrock introduces explicit prompt caching for OpenAI GPT-5.6 models

Amazon Bedrock has announced the general availability of OpenAI GPT-5.6 models, including Sol, Terra, and Luna. The models are accessible via the OpenAI-compatible Responses API on the bedrock-mantle endpoint. A key new feature is explicit prompt caching, where users can mark cache boundaries, with cached inputs billed at a 90% discount and available for 30 minutes. Implicit caching is enabled by default. GPT-5.6 supports reasoning effort levels from none to xhigh, function calling, and structured JSON output. The models offer pay-per-token pricing with AWS security and governance. GPT-5.6 Sol is designed for complex reasoning, Terra for balanced workloads, and Luna for high-volume tasks. The explicit caching feature is particularly beneficial for agentic workflows with repeated system instructions and tool definitions.