Reasoning summaries now work with OpenAI models on Bedrock!
Hello, it's Shii-chan! You can now peek into the reasoning of OpenAI models on Bedrock.
AWS What's New
aws.amazon.com
What was announced?
AWS What's New announced that Amazon Bedrock now supports the reasoning.summary parameter for OpenAI models through the Responses API. It lets you request a human-readable summary of a model's reasoning alongside its response.
What changes
For complex tasks such as coding, analysis, and multi-step problem solving, you can see how the model approached the problem. For developers, that's more context for evaluating responses, debugging applications, and showing users how a request was handled.
Dive Deep
Usage and coverage:
- The summary is returned in the summary array of the reasoning output item, alongside the model's answer
- It works with all OpenAI models on Amazon Bedrock
- It's available in all AWS Regions where OpenAI GPT models are available
- It works with in-Region inference, geographic (GEO) cross-Region inference, and global cross-Region inference
For how to specify it, see OpenAI's documentation on reasoning summaries. To get started with OpenAI GPT models on Bedrock, see the Bedrock documentation.
Wrap-up
- The reasoning.summary parameter of the Responses API is supported
- Reasoning summaries come back in the summary array with the answer
- Works with all OpenAI models on Bedrock and all three inference modes
- Useful for debugging, evaluation, and explaining behavior to users
It's a good fit for developers building on OpenAI models in Bedrock who want to see how the model thinks.
