Every token, counted once.
Records are deduplicated by response id, streamed chunks are never summed, and cost is kept in integer micro-dollars.
Log records arrive one by one. Records that repeat a response id are struck out and skipped. Token totals by category and the spend update as records are counted.
Today
- Input
- 123,364
- Cache read
- 5,535,799
- Cache write, 5 min
- 192,444
- Cache write, 1 hour
- 77,392
- Output
- 117,579
- of which reasoning
- 23,872
Spend
$4.31partial
Partly unpriced, cost leaves it out
Prices from LiteLLM and models.dev















































