Track usage

Usage and cost show what your traffic is actually consuming. It’s the reporting side of usage limits: limits set the cap, usage shows where you stand against it.

​

Where to see it

• The Dashboard shows the whole deployment: estimated spend, tokens, requests, how much traffic was rerouted and what that saved, spend over time, and the top groups, keys, and models by spend.
• The Leaderboard ranks spend, with a tab for API keys, Groups, Users, Providers, and Models, so you can see where cost is coming from without opening keys one at a time.
• The Usage limits page shows each limit’s current usage against its amount for the active window.
• A key’s detail page shows requests, tokens, and estimated cost over time, broken down by model, provider, or token type.

​

What’s tracked

For every key, the gateway records each day (UTC):

• Requests: how many calls the key made.
• Tokens, split into the five types that price differently:
  • Uncached input and Cached input, the prompt tokens at standard and cache-read rates.
  • Cache write, the input tokens charged for populating the provider’s cache.
  • Non-reasoning output and Reasoning output, so you can see how much of a bill is a reasoning model thinking.
• Estimated cost: in USD, and split across those five token types.
• Routing: which incoming model and provider the caller asked for, and which target model and provider served it.

Usage can be sliced by day, by key, or by model. On the spend chart, switch between Spend and Tokens, and break either down by Model, Provider, Token type, or Direct/Rerouted.

​

Rerouting savings

When a routing rule sends a request to a different model than the caller asked for, the gateway prices the same tokens twice: once at the model that actually served the request, and once at the model the caller named. The difference is the saving from that reroute, reported as Saved by routing alongside Rerouted, the share of requests a rule redirected. A key’s detail page also breaks this out as a rerouting flow, showing which requested model ended up on which serving model. A reroute onto a more expensive model shows as a negative saving, which is the signal that a rule is costing you money rather than saving it.

​

How cost is estimated

The gateway multiplies token counts by its cached per-model pricing, using the provider’s separate rates for cached input where they apply. Treat it as a close estimate for tracking and budgeting, not as a replacement for your provider’s invoice.

​

Go deeper

Usage answers “how much.” For request-level detail, including inputs and outputs, latency, and the trace of each call, see Observability.