Saved in the last 30 days
$0
Would have cost
Actually cost
Monthly pace
Requests
Attacks blocked
Avg decision time
Where requests went
Each request is blocked, sent to the cheap model, sent to the strong model, or answered from cache.
Quality check
Were the cheap model's answers good enough? A sample is re-asked to the strong model, which judges them.
Money saved per day
Estimated savings compared with sending every request to the strong model.
Show as table
No traffic yet
Point your app at this server and send a request with model="auto" to /v1/chat/completions.
Savings will show up here.