Skip to main content
A model’s processing location is where it processes the text you send it. Every hosted model uses one of two locations:
  • global means any of the provider’s regions, including regions outside the United States.
  • us means only regions in the United States.
Some models offer both, so you choose one with processing_location when you declare an embedded field.

​
Locations by model

The Processing location column of the model gallery lists each model’s location. Currently only text-embedding-3-small and text-embedding-3-large let you choose, between global (the default) and us. gemini-embedding-2 processes in global and accepts processing_location only when it’s set to global. Every other model rejects it.

​
Set the processing location

processing_location is a property of an embedded field on a document index, set on embed next to dimension, so it requires API version 2026-07. One value covers both stored text and queries. To keep both in the US, set it to us:
US processing
These rules apply:
  • The processing location is currently fixed at creation, like the rest of the embedding configuration. When you describe the index, the field’s embed.processing_location shows the location the model uses now.
  • On the Starter and Builder plans, text-embedding-3-small and text-embedding-3-large are limited to global. On the Standard and Enterprise plans, you can set processing_location to us.
  • If your organization moves to the Starter or Builder plan, an index created with us keeps working but processes text in global, and describing or listing the index shows global. If you upgrade again, the index goes back to us.
  • A location the model doesn’t currently offer, such as eu, is rejected with a 400.
  • The List models operation lists processing_location, with its accepted values, in a model’s supported_parameters on API version 2026-07 and later.