OutYet reporting
AWS adds GPT-5.6 Sol to Bedrock's Responses API
AWS has documented GPT-5.6 Sol in Amazon Bedrock, pairing the flagship model with Bedrock-specific caching, regional controls, and AWS security boundaries rather than a separate model variant.
AWS published a July 13 guide saying that GPT-5.6 Sol, Terra, and Luna are available through Amazon Bedrock's Responses API. For Sol, AWS lists US East (N. Virginia) and US East (Ohio), while Terra and Luna also include US West (Oregon). This is a deployment-surface change for teams already standardized on Bedrock: they can use the OpenAI family through AWS's control plane rather than treating the model as an OpenAI-only API dependency.
The AWS publication follows OpenAI's stated rollout sequence. OpenAI described Sol, Terra, and Luna as a limited preview on June 26, then announced general availability of the GPT-5.6 family on July 9. The provider describes Sol as the flagship tier, Terra as a lower-cost tier aimed at everyday work, and Luna as the fastest, least expensive tier. AWS's guide therefore documents an additional hosted route after the provider's own rollout, rather than describing a distinct successor to Sol.
OpenAI's July announcement makes several performance and price comparisons with GPT-5.5 and competing models, including its own benchmark tables and claims about token use, latency, and estimated cost. Those figures are useful for understanding OpenAI's positioning, but they are provider-reported comparisons rather than an independent procurement result. The practical distinction is that Sol is the model OpenAI positions for the most demanding reasoning workloads, while Terra and Luna are the intended alternatives when a workload can trade peak capability for lower cost or faster inference.
For Bedrock users, the more concrete operational details are prompt caching, regional placement, and data-handling constraints. AWS says developers can set explicit cache breakpoints, receive a 90 percent cached-input discount, and reuse cached input for at least 30 minutes. It also says classifier-flagged traffic may be retained for automated abuse detection for up to 30 days. Teams should test the supported region, cache behavior, IAM and data-governance requirements, and task-level quality before moving an existing production workflow to the new route.