Skip to navigation
Model VaultDeploy & manage

Regions

You choose a region when you create a vault. Every model in the vault runs in that region, and your prompts and responses are processed there. A vault’s region can’t be changed after creation.

Available regions

RegionShown in the app asStandard VaultEncrypted Vault
United StatesGlobalAvailableAvailable
European UnionEUAvailableComing soon
CanadaCANAvailableComing soon

Model Vault Encrypted currently runs in the United States. Regional serving for Encrypted Vaults in the European Union and Canada is coming soon.

Model availability by region

Each region runs a specific set of GPUs, so the models and GPU options you can deploy depend on the region:

  • United States: every supported model and GPU.
  • European Union: models that run on NVIDIA B200 or NVIDIA L4 GPUs.
  • Canada: models that run on NVIDIA GB200 or NVIDIA L4 GPUs. GB200 is available through Cohere sales.

When you create a vault, the model catalogue only shows the models and GPUs available in the region you selected. For the full model lists, see Standard Vault supported models and Encrypted Vault supported models.

Data residency

Choosing a region helps you meet data residency and regulatory requirements: inference for your vault runs on GPU clusters in that region, and Zero Data Retention means prompts and responses are not retained after a request completes. For compliance details specific to Encrypted Vaults, see Compliance.