Databricks Model Serving Cost Monitoring
Open AI Services in the sidebar to investigate AI spend. The page covers Databricks serving endpoints and other AI products when system.serving and billing data are available. Genie has its own page; see Genie.
Overview
Section titled “Overview”Shows AI spend by product: daily spend, each product’s cost and share, and the open AI insights.
Model Serving
Section titled “Model Serving”Shows the trend of the top endpoints and the endpoint list. Each row shows the endpoint, its type, requests, tokens, and cost. Select an endpoint to see its top requesters.
Databricks reports requests only for endpoints with usage tracking enabled. An endpoint without it shows 0 requests even when it serves traffic.
Efficiency
Section titled “Efficiency”Shows the cost per 1,000 tokens by endpoint, and zombie endpoints with their last request, days inactive, and cost.
Vector Search
Section titled “Vector Search”Shows Vector Search endpoints with their usage, cost, and daily spend.
Filters
Section titled “Filters”Use global filters for time range, workspace, organization, tags, and cost mode.
Waste detection
Section titled “Waste detection”A Zombie Endpoint insight flags a serving endpoint that used to receive requests, has had none for at least 30 days, and still costs money. Check that request tracking is complete, and confirm ownership, before you stop the endpoint.