We tried this, but the quota for Opus models defaults to 0 on VertexAI and quota increase requests are auto-rejected.

Any tips?