Amazon Bedrock Now Lets OpenAI Models Search the Web!
Hey everyone, I'm Shii-chan! Today I've got a handy new feature for Amazon Bedrock — OpenAI models can now search the web without ever leaving AWS!
AWS What's NewWhat was announced?
According to AWS's What's New, Web Search on Amazon Bedrock is now generally available (GA). It's a built-in, server-side tool that runs entirely within AWS, letting OpenAI models (GPT-5.4, GPT-5.5, and GPT-5.6 Sol/Terra/Luna) ground their responses with current web knowledge. The search itself happens inside your secured AWS environment, with zero data egress.
The story so far
Until now, adding web grounding meant onboarding a third-party search provider, managing a separate API key and billing relationship, building custom orchestration, and running an extra compliance review for each external vendor. That's a lot of heavy lifting just to add a bit of web search.
What changes
With Web Search, you can turn on grounding with a single parameter added to an existing API call. No vendor onboarding, no external API to orchestrate, and no extra vendor security reviews. It integrates through a standardized tool-use interface compatible with the OpenAI Responses API, and Bedrock handles the entire search lifecycle server-side, returning a grounded, cited response in one API call.
Dive Deep
Web Search draws on a web index operated by Amazon, spanning tens of billions of documents that are refreshed continually, combined with a built-in knowledge graph for verified facts. It's built on experience from Alexa+, Amazon Quick, and Kiro.
Rather than returning raw pages, it performs semantic snippet extraction, delivering context-efficient results sized for the model's context window, with low latency.
Currently, Web Search is available in these regions:
- US East (N. Virginia)
- US East (Ohio)
- US West (Oregon)
For pricing and setup details, check the AWS technical blog, the Amazon Bedrock User Guide, and the pricing page.
Wrap-up
- Web Search on Amazon Bedrock is now GA, for GPT-5.4, GPT-5.5, and GPT-5.6 (Sol/Terra/Luna)
- Runs entirely within AWS with zero data egress for web grounding
- Enabled with a single added parameter on an existing API call, no third-party search provider onboarding required
- Combines Amazon's web index and knowledge graph, with semantic snippet extraction for low-latency results
- Available today in US East (N. Virginia, Ohio) and US West (Oregon)
This one's for engineers running OpenAI models on Bedrock who want easy, up-to-date web grounding without bringing in an external search vendor!