Token optimization savings

Text we strip out before a call reaches the model — bulky tool output, old chat history, long answers. Everything here is what you did not have to pay for.
BetaMeasured + estimated
Tokens saved
18.7B
19.7% of 94.7B you would have sent
Cost saved
$34,346.20
19.7% of $174,346.20, in this window
Tokens
76.1B sent · 18.7B saved
Cost
$140,000.00 sent · $34,346.20 saved

Sent Saved (not sent)

  • Agents optimized252 of 318
  • Calls through it5,464,745 of 8,795,665
  • Token optimization252 of 318
  • Minimal-code mode0 of 318
  • Prompt comments0 of 318

Chat history

History selection: the messages a request could carry against what it sent
Without optimization
26.4B
Sent
21.2B
Saved
5.2B · 19.7%
Messages dropped
15,356,952
Sent Saved (not sent)Oct 15 – Nov 14 (UTC)

Based on 3,791,232 of 4,206,110 requests (90%). About $5,926.27 of input not billed. Fewer tokens, not necessarily better answers.