About Foundry Map

A reference for Azure AI Foundry model availability — which models are deployable in which regions, for which deployment type, and at what price. Free, open, updated daily.

Why this exists

The official Azure docs spread Foundry information across a dozen pages: per-model availability tables that go stale, retirement notices on a separate page, deployment-type definitions on a third, pricing on a fourth. Answering a basic question — "can I run gpt-5 in switzerlandnorth on Provisioned today, and how much will it cost?" — means cross-referencing multiple sources and hoping none of them are out of date.

Foundry Map is a static site rebuilt daily from Microsoft's own ARM and Retail Prices APIs, so the answer to that question is one page-load away.

How it works

A Python pipeline calls Microsoft.CognitiveServices/locations/{location}/models for every Azure region that hosts AI Foundry, merges the result with the Azure Retail Prices API, and commits the normalised JSON to this repo. The static site you're reading is rebuilt on every commit.

No accounts, no tracking beyond privacy-friendly Plausible analytics (no cookies, no PII), no JavaScript framework runtime — just static HTML over the data.

Data freshness

The pipeline runs once a day in the morning, refreshing models, regions, and prices from the Azure ARM and Retail Prices APIs. The footer of every page shows the timestamp the data was last refreshed.

Learn the concepts

Two of Foundry's foundational concepts get their own deep-dive on the Learn section — written with analogies that are meant to stick:

• Deployment types — Standard, Provisioned, Batch crossed with Single region, Data Zone, Global, plus Priority, Serverless, and Managed Compute. Restaurant analogy, decision framework, PTU break-even.
• Model lifecycle — Preview, GA, Retiring, Deprecated, Unknown. Broadway analogy, notice windows, ARM enum mapping, operational decision rules.

Known limitations

• Price coverage is best-effort (~13%). Retail Prices meter names use an ad-hoc encoding per model family. Expanding coverage is a roadmap item.
• Anthropic, Mistral, Cohere, and some other Marketplace-billed models don't appear in the Retail Prices API at all.
• Default quota (TPM/RPM) isn't surfaced yet — pending a stable source.
• Fine-tuning availability is out of scope for MVP.

Contact

Spotted a bug or have a feature request? Reach out via waynegoosen.com.

Disclaimers

Not affiliated with Microsoft. "Azure", "Microsoft Foundry", and related names are trademarks of Microsoft Corporation; used here descriptively.

No warranty. Data is provided as-is from Microsoft's published APIs. Pricing, availability, and lifecycle status can change without notice — verify with the Azure portal, the Retail Prices API, and the Foundry retirements page before designing production migrations or commercial commitments around anything you read here. Numbers in the Learn deep-dives reflect Microsoft's published commitments at the time of writing and may have changed since.