OutYet reporting
GPT-5.6 Sol reaches Amazon Bedrock, with regional limits still in view
AWS has added OpenAI's GPT-5.6 family to Bedrock, giving existing AWS customers a new deployment route while leaving performance and operational questions for their own evaluation.
Amazon Web Services said on July 13 that GPT-5.6 Sol, Terra, and Luna are generally available through Amazon Bedrock. AWS says the models can be called through the Responses API, that Bedrock pricing matches OpenAI's first-party rates, and that usage can count toward existing AWS commitments. For teams already operating on Bedrock, this is principally a hosting and procurement option: it can place the GPT-5.6 family inside an existing AWS account and control plane, but the announcement does not establish a different underlying model capability profile.
The Bedrock availability follows OpenAI's July 9 introduction of the GPT-5.6 family across its own ChatGPT, Codex, and API surfaces. OpenAI describes Sol as the flagship tier, Terra as a lower-cost tier positioned against GPT-5.5, and Luna as the fastest and most affordable tier. Its published API rates are $5 input and $30 output per million tokens for Sol, $2.50 and $15 for Terra, and $1 and $6 for Luna, so the family is designed to let developers trade capability, speed, and token cost by workload rather than use one model for every request.
OpenAI reports sizable improvements for Sol on its selected coding, computer-use, cybersecurity, and science evaluations, including comparisons with GPT-5.5 and competing frontier systems. Those comparisons are useful as vendor-reported launch evidence, but they are not a substitute for a deployment-specific test: the published figures use particular reasoning settings, benchmarks, and cost assumptions. Technical buyers should therefore treat the three-tier lineup as a routing hypothesis, then measure task success, latency, token use, and failure behavior on their own tools and data before moving a production workflow.
The immediate constraint is geography. AWS says Sol is initially available in US East (N. Virginia) and US East (Ohio), while Terra and Luna are also offered in US West (Oregon). Teams with regional residency requirements, multi-region failover plans, or strict latency targets should verify that footprint before standardizing on Sol. The cited announcements confirm the new Bedrock path and describe the model tiers, but they do not replace workload-level validation of quotas, latency under load, tool behavior, or the practical cost of higher reasoning settings.