# Amazon Bedrock adds reasoning summaries for OpenAI models > AWS says Bedrock developers can request reasoning summaries for OpenAI models, with support for in-Region and cross-Region inference. Canonical URL: https://www.devobs.io/news/news-event-2564673af1baaed81c3d8d9a/ By: Sofia Reyes Published: 2026-10-11T04:39:53.061Z Updated: 2026-10-11T04:39:53.061Z Event date: 2026-10-09 Section: AI Amazon Web Services announced on October 9, 2026, that Amazon Bedrock supports the `reasoning.summary` parameter for OpenAI models through the Responses API. According to the [AWS announcement](https://aws.amazon.com/about-aws/whats-new/2026/10/amazon-bedrock-reasoning-summaries-openai/), developers can request a human-readable summary of a model’s reasoning alongside its answer. AWS says the additional context can help developers evaluate responses, debug applications and show users how a model approached tasks such as coding, analysis and multi-step problem solving. ## Where the summary appears AWS says the summary is returned in the `summary` array of the reasoning output item. That gives developers a defined location from which to retrieve the explanatory content. For application teams, a useful integration choice is to process and display summaries separately from answers. Keeping the two distinct can help prevent downstream consumers from treating explanatory commentary as the requested result. ## Availability and regional limits According to the [announcement](https://aws.amazon.com/about-aws/whats-new/2026/10/amazon-bedrock-reasoning-summaries-openai/), support is available for all OpenAI models on Amazon Bedrock in all AWS Regions where OpenAI GPT models are available. AWS explicitly includes three inference options: - In-Region inference - Geographic (GEO) cross-Region inference - Global cross-Region inference This is not a claim of availability in every AWS Region: the stated boundary remains the regions where OpenAI GPT models are available. ## Engineering implications: context, not validation Treat a readable explanation as context rather than proof that an answer is correct. Coding assistants should still validate generated changes with tests, and analytical applications should still check results against underlying data. Summaries should supplement those checks, not replace them. Before rollout, confirm the supported request settings and assess any cost or latency impact for the intended workload. AWS’s announcement does not list accepted parameter values, provide an implementation example or quantify pricing and latency effects. ## Source references -