About Bedrock Explorer

An independent reference for Amazon Bedrock: which model runs in which AWS region, at what price, with which capabilities, and where the inference physically executes. Free, no account, no tracking beyond aggregate analytics, not affiliated with Amazon Web Services.

Models tracked: 116
AWS regions probed: 31
Jurisdictions assessed: 25
Snapshot: Aug 27, 2026

Who maintains it

Dorian Richard, Cloud & GenAI engineer. I build generative-AI systems on AWS for a living — the kind where somebody eventually asks what the monthly bill will be, whether the model is available in Frankfurt, and whether the data leaves the country. This site started as the spreadsheet I kept re-making to answer those three questions, and it exists because the answers are genuinely hard to assemble from the AWS console and pricing pages.

I work at The Intelligence Partners, and post on LinkedIn when the data turns up something worth arguing about.

Where the numbers come from

A pipeline queries the AWS Bedrock, Price List and Service Quotas APIs across every commercial region where Bedrock is offered, and writes one versioned snapshot. The site renders that snapshot and never calls AWS itself, so what you see is reproducible and auditable — the snapshot is downloadable in full.

Where AWS publishes nothing — modern Anthropic prices, context windows, tool-use support — the gap is filled by a hand-verified overlay that is tagged as such in the data and badged in the UI, never blended silently into API figures. A model with neither an API price nor a verified one shows "not published" rather than a guess. The methodology page is the long version, including what the site deliberately does not cover.

How it is funded, and the conflict of interest

Nobody pays for placement, there are no ads, no affiliate links and no sponsored models. The site is free because it is useful to me too.

The one commercial interest is disclosed on every page it appears on: some pages carry a link to my consultancy, for readers who reach a point where the table stops being enough. That is the whole business model, and it is the reason the data has to stay disinterested — a reference that shaded its numbers toward a sales pitch would be worth nothing to anyone, including me.

Corrections

If a number here is wrong, I want to know — a wrong price on a reference site propagates into somebody's budget. Tell me what you found and I will fix the pipeline rather than the page, so it stays fixed.