AWS has announced that OpenAI’s GPT-5.6 Sol, Terra, and Luna models are now generally available on Amazon Bedrock. The release is paired with explicit prompt caching, a feature that lets users specify which portions of a prompt should be cached and reused across requests.
According to the post, the accompanying material explains how to get started, configure explicit caching, and migrate existing GPT workloads. The stated goal of the caching capability is to reduce inference cost.
Why it matters
Explicit control over cached prompt content targets a common cost driver in running large models: repeatedly processing the same prompt segments. By reusing cached parts, teams can lower inference expenses on recurring workloads.
Who should care
Teams already running GPT workloads on Amazon Bedrock, or planning to migrate them, are the primary audience, particularly those focused on managing inference costs.