Support Prompt Caching with Vertex

[willriley]

To my knowledge, prompt caching isn't supported when using Claude on Vertex, either via the messages API or the SDKs. Is there any ETA on when that will be added?

[jennmueng]

We use Claude through vertex ai too, keeping an eye on this.

[ggdupont]

That's a bummer for us and we might need to switch model if this is not supported soon.

[codenprogressive]

+1 Any updates on enabling context caching for Claude available through VertexAI?

[daivitran]

It is supported here: https://cloud.google.com/vertex-ai/generative-ai/docs/partner-models/claude-prompt-caching

[sammcj]

This is absolutely critical for things like agentic coding, where prompt caching results in around a 98% saving in cost.

Prompt caching works really well with Claude models GCP VertexAI directly, so this will be a bit of a blocker for a lot of folks wanting to use Claude on GCP, or using Claude models and wondering why they're paying 98% more.