How to turn on fast mode in Claude Code on Google Cloud Vertex AI

Starting with Claude Code 2.1.322, fast mode also works when you connect through Google Cloud Vertex AI. The changelog lists three regions: global, us and eu.

1. Turn on the Vertex integration and set the region to one of the three.

export CLAUDE_CODE_USE_VERTEX=1
export CLOUD_ML_REGION=global
export ANTHROPIC_VERTEX_PROJECT_ID=YOUR-PROJECT-ID
claude

1. In the session, type /fast, press Space to turn it on, then press Enter.

If Vertex refuses the fast mode request, Claude Code falls back to standard speed and keeps answering. Fast mode costs more per token, so turn it on only for work where speed matters. The fast mode docs still say it isn't available on Vertex; the support appears only in the changelog.