Token meter

Flattening the raw FHIR bundle into tabular rows is the whole trick. The model reads a fraction of the context, so a sweep costs cents instead of dollars, and the Evals show the scores hold on that smaller context. The bars below put the two side by side for this run.

Token meter

Context tokens and cost for this run, measured on the flattened context Clawback actually reads.

replayed run

92.1%

smaller context after flattening

3924

input tokens

10243

output tokens

$0.17

cost this run

Eval scores are measured on the flattened context. Latest anomaly F1 is 100.0%. A raw-context A/B baseline has not been run yet.

see Evals

Cost is computed at the published claude-sonnet-5 rates of $3.00 per million input tokens and $15.00 per million output tokens. Token counts are estimated at about 4 characters per token. A live API run measures them exactly.

View as table

raw FHIR JSON tokens	24610
flattened tokens	1936
reduction	92.1%
input tokens	3924
output tokens	10243
cost this run	$0.17

The chip on the meter marks whether these numbers come from a replayed run or a live one. Replay is the default so the demo is always fast and free. Live mode runs your own key, and the meter then shows the real usage and cost of that run.