title: AI Usage description: Two dashboards on what the agent spent: Tokens & Cost rolls up by money, AI Models rolls up by model, provider, and streaming latency.
The AI Usage group covers two dashboards that answer the question what did the agent spend in token and economic terms - one tab per perspective. Both pages read from the same trace stream documented in Observability Architecture, and both feed off gen_ai.* span attributes that Spring AI emits per OpenTelemetry GenAI semantic conventions.
flowchart LR
Span["gen_ai.client.operation<br/>span"]
Price["Per-model<br/>rates"]
Curr["Display<br/>currency"]
TC["Tokens & Cost"]
AM["AI Models"]
Span --> TC
Span --> AM
Price --> TC
Curr --> TC
Per-model rates and Display currency are both configured exclusively through the Model Pricing Manager dialog opened from the Tokens & Cost dashboard - the dialog is the only supported edit surface.
The two tabs share the same span data - they slice it differently. Tokens & Cost rolls up by money and tokens; AI Models rolls up by model identity, provider, and latency characteristic (including streaming TTFT). Cost lives only on Tokens & Cost because pricing is a per-model lookup applied at read time, not a property of the trace itself.
For the underlying pipeline that captures these spans, see Observability Architecture → The pipeline. For the Model Pricing Manager dialog and how cost is computed, see the Tokens & Cost page.