OpenAI GPT-5.6 Sol, Terra, and Luna are now generally available on Amazon Bedrock, accessible via the OpenAI Responses API on the bedrock-mantle endpoint. Sol targets autonomous coding and deep reasoning, Terra balances performance and cost for production workloads, and Luna handles high-volume, latency-sensitive tasks. All three support a 272K-token context window, tool calling, and configurable reasoning effort levels. The post covers authentication setup using the BedrockOpenAI client, prompt caching (both implicit and explicit modes with a 90% discount on cached tokens), connecting the OpenAI Codex coding agent, and managing quotas with exponential backoff for throttling.
Table of contents
The GPT-5.6 family on Amazon BedrockAccessing GPT-5.6 through the bedrock-mantle endpointGet started with GPT-5.6 on Amazon BedrockReduce cost with GPT-5.6 prompt cachingUse GPT-5.6 with CodexQuotas and scalingClean upConclusionAbout the authors31 Impressions