How do I track usage and spend per model?

Three ways, from quickest to most powerful:

Dashboard — click Model APIs in the sidebar for per-model views.

Usage endpoint —GET /v1/model_apis/usagereturns request counts and token counts (input, cached input, uncached input, output) in 1-minute, 1-hour, or 1-day buckets, grouped or filtered by API key, user, or model. This is the tool for "which key/teammate/model is spending the money."

Prometheus export — baseten_model_api_tokens_total and baseten_model_api_inference_requests_totalstream the same numbers into your own observability stack.

Note: usage data is available from August 5, 2026 onward — earlier Model API usage wasn't backfilled.

Docs: Usage · Metrics export