Introducing regional inference routing on Boundless: choose where your inference runs

Boundless now runs inference in the region you choose: the US, the EU or Asia, available now on request, with other regions set up when you ask. Regional inference routing gives teams with data residency requirements a direct answer to the first question a security review asks about an AI vendor: where requests are processed.

Zero data retention (ZDR) is built in on the open models Boundless serves, so prompts and outputs aren’t stored in any region. Access is set up by request.

Security reviews start with where data goes

Before a company sends prompts to an AI vendor, its security, legal or procurement team asks two things: where the data is processed, and whether it's kept. For companies in the EU and Asia, the first answer often has to be a specific region, because an internal policy or a customer contract says so.

Many inference APIs don't let you choose, so requests can run outside the region a policy requires. The team then negotiates an exception or skips the service, and running open models in-region usually means buying GPUs there and operating the serving stack in-house.

Choose a region, and requests are processed there

Regional inference routing, also called inference residency or data residency for AI, lets a team run its workload in the US, the EU or Asia. Regional access runs on a separate console and API endpoint, with its own team account and API key. Approved EU teams, for example, sign in at eu.inference.boundless.network and send requests to api.eu.inference.boundless.network/v1, so requests are processed in the EU. Any open model Boundless serves can run in any of these regions, and a region that isn't listed yet is available on request.

Zero data retention in every region

Where a request runs is half of what a review asks, and the other half is what happens to it afterwards. Boundless keeps zero data retention on the open models it serves, so prompts and outputs aren't stored or used for training in any region. Billing keeps only token counts, the model, a timestamp and the cost.

Together, the two answers fit on one line of a vendor review: processed in the region you chose, and not stored.

Built for teams with data residency requirements

Regional inference routing is for companies in the EU and Asia whose policies require in-region processing, and for AI products selling to enterprises that write data location into their contracts. It also opens hosted inference to teams that avoided it for this reason, since keeping requests in a region no longer means running GPUs there.

Request a region

Send the Boundless team your target region, your workload and the models you want to run there, and the team confirms pricing and setup. In the EU, for example, DeepSeek V4.1 Flash 1017 and its abliterated version for authorized security testing run at the same per-token prices as the main platform. The full catalog is on the models page, with per-token rates on the pricing page.

Teams already on Boundless can request a region from the inference console: open Settings, select Regions, and pick the region you need.

Get started on Boundless

For teams that want high-performance inference that keeps getting faster, Boundless serves open models through one API, with zero data retention built in. Get started today.

Frequently asked questions

What is regional inference routing?

Regional inference routing lets you choose the region where an AI provider processes your inference requests. On Boundless, that means the US, the EU or Asia, with zero data retention on the open models Boundless serves.

Which regions does Boundless support?

The US, the EU and Asia are available now on request. Boundless sets up other regions when a team asks.

Does Boundless store prompts or outputs?

No, zero data retention applies to the open models Boundless serves in every region, so prompts and outputs aren't stored or used for training. Billing keeps only token counts, the model, a timestamp and the cost.

Does my current API key work in another region?

No, regional access uses its own console, API endpoint and API key, which the Boundless team sets up with you.

Which models can I run in each region?

Any open model Boundless serves can run in any region on request, and the Boundless team confirms pricing and setup for yours. In the EU, for example, DeepSeek V4.1 Flash 1017 and its abliterated version run at the same per-token prices as the main platform.