AWS announced the general availability of OpenAI’s GPT-6 Sol and GPT-6 Luna in Amazon Bedrock on September 22. Sol targets recurring complex work, including software development and multistep processes across tools, while Luna is intended for focused, repeatable tasks at high volume. Customers can access both through the Bedrock console or supported APIs.
According to AWS, GPT-6 Sol can implement features, debug problems, refactor and review code, analyze data, and carry out workflows spanning multiple tools and applications. AWS says improvements over GPT-5.6 Sol help the model preserve the reasoning context behind its decisions as work moves from investigation to implementation and validation.
AWS also cites an internal OpenAI factuality evaluation in which GPT-6 Sol made approximately half as many factual mistakes as GPT-5.6 Sol. This is a vendor-reported internal result rather than an independent benchmark. The post provides no evaluation methodology, task set or absolute error totals, limiting direct comparisons with production workloads.
GPT-6 Luna is positioned for extracting information from large document collections, summarizing incoming material, classifying inputs and answering focused questions. Developers can adjust reasoning effort for each request to balance quality, responsiveness and cost. AWS reports improved factual reliability in OpenAI’s evaluations but gives no numerical Luna results in the post.
Both models support explicit prompt caching in Bedrock. Applications can mark instructions, tool definitions, policies or reference material for reuse, allowing later requests to concentrate processing on new input. AWS identifies coding assistants, policy-based support systems and document pipelines using a consistent extraction schema as potential uses.
Practical context: The product split supports a tiered workflow: Luna can classify incoming requests, Sol can investigate harder cases, and GPT-6 Astra can be reserved for decisions where greater reasoning depth may matter. Such routing could reduce unnecessary computation, but its practical benefit will depend on pricing, latency, reusable context and routing quality. AWS says Sol and Luna have significantly lower API pricing than their GPT-5.6 predecessors; current rates and regional support must be checked in the Bedrock documentation.
Bedrock provides access controls through AWS IAM, invocation auditing through CloudTrail and private VPC endpoints through AWS PrivateLink. AWS says inference data is not used for model training and customers do not have to share it with OpenAI. For automated abuse detection, AWS retains classifier-flagged traffic for up to 30 days and processes it programmatically. Customers can request zero data retention through their AWS account team.
| Model | Suggested role | Example tasks |
|---|---|---|
| GPT-6 Luna | High-volume focused operations | Classification, data extraction and summarization |
| GPT-6 Sol | Recurring complex tasks | Code development and review, data analysis and tool-based workflows |
| GPT-6 Astra | Work requiring the highest result quality | Cases where additional reasoning depth may materially change a decision |
Sources
Event date: 2026-09-22. Primary source date: 2026-09-22.
