Get started with OpenAI GPT-5.6 Sol, Terra, and Luna on Amazon Bedrock

OpenAI GPT-5.6 Sol, Terra, and Luna are now generally available on Amazon Bedrock. Teams can pick among the three models and run them inside the Bedrock surface instead of wiring a separate OpenAI path for those workloads.

Inference goes through the Responses API on the bedrock-mantle endpoint. The same guidance covers model selection, prompt caching to cut cost on repeated context, wiring the OpenAI Codex coding agent, and planning for quotas and scale.

For engineers, this puts Sol, Terra, and Luna behind Bedrock’s existing access, auth, and ops model. Builders who already route traffic through Bedrock can add these models without a second provider stack for inference, caching, or agent-linked coding flows.

The move places OpenAI’s GPT-5.6 Sol, Terra, and Luna alongside other foundation models already sold through Bedrock. That matters for shops standardizing on one cloud ML control plane rather than juggling multiple vendor endpoints for model choice and agent tooling.

Practical next steps: choose Sol, Terra, or Luna for the task; call the Responses API on bedrock-mantle; turn on prompt caching where prompts repeat; connect Codex if you want the coding agent path; and size quotas and scale before production load hits the new models.