Regions
Choose whether an API key's inference runs in Europe, the United States, or both.
Inference can run in Europe, the United States, or both. The setting is per API key. Different keys on the same account can use different regions.
Region selection controls where inference runs, not how requests are routed. All requests are routed through Europe, regardless of a key's region setting.
Set regions for a key
Open Account → API keys and select the regions a key may use:
- Europe
- United States
New keys allow every region by default. Uncheck a region to keep that key's inference in the other.
How the region setting is applied
- The requested model is limited to deployments in the key's allowed regions, then a healthy deployment is selected from the matching set.
- If the requested model is not served in any allowed region, the request fails rather than running inference in a region the key does not allow.
- Client-supplied region routing hints are ignored. The key's configuration is authoritative.
Availability
A model can be available in one region and not another at a given moment. The account page shows which models are currently available in each region. See Models for the full catalog.
Related
- Key controls: quotas, expiry, and revocation.
- Privacy: how requests are processed and retained.