Skip to content
BENIBIX Learn

CLOUD_ML_REGION vs VERTEX_REGION_CLAUDE_*

People think CLOUD_ML_REGION and the VERTEX_REGION_CLAUDE_ variables do the same job. One sets the region for every model. The other sets it for one model. Claude Code checks the model's own variable first. If it is not set, it uses CLOUD_ML_REGION, and then us-east5. Set CLOUD_ML_REGION once. Add a model's own only when that model needs a different region.

AI⌨️ Claude Code

CLOUD_ML_REGION

It sets the Google Cloud region Claude Code uses.

CLOUD_ML_REGION sets the Google Cloud region Claude Code uses on Agent Platform. It can be global, a multi-region such as eu, or one region. If it is not set, or not shaped like a region, Claude Code uses us-east5. A model can have its own region that wins over this one.

Example
Emma sets this to global, so Google picks a region with room. One model has no global spot, so she gives it its own region.
Why it matters
The region decides which models you can reach and where your data is handled.
Full entry
AI⌨️ Claude Code

VERTEX_REGION_CLAUDE_* Variables

It sets the Google Cloud region for one model.

A VERTEX_REGION_CLAUDE_ variable sets the Google Cloud region for one model. Most models have one, such as the one for Opus 5.5. It wins over CLOUD_ML_REGION for that model only. Use it when your main region is global but a model is not.

Example
Jack runs on global, but one old model is not. He sets that model's own region, and it works while the rest stay on global.
Why it matters
One model with no global spot would fail without it, while the rest work.
Full entry

In one line

CLOUD_ML_REGION
It sets the Google Cloud region Claude Code uses.
VERTEX_REGION_CLAUDE_* Variables
It sets the Google Cloud region for one model.