WEBVTT

00:00:00.000 --> 00:00:04.894
This implementation explicitly includes six hundred tool-result tokens and four hundred output

00:00:04.894 --> 00:00:09.787
tokens per step. Starting with a two thousand token prompt, full-history inputs

00:00:09.787 --> 00:00:14.681
are two thousand, three thousand, and so on through nine thousand for

00:00:14.681 --> 00:00:19.574
step eight. Their sum is forty four thousand. Output totals thirty two

00:00:19.574 --> 00:00:24.468
hundred. At point zero zero six pixels per modeled token the bars

00:00:24.468 --> 00:00:29.361
span two hundred sixty four and nineteen point two. This differs from

00:00:29.361 --> 00:00:34.255
the earlier calculator with a fixed fifty-token overhead. These are estimated tokens

00:00:34.255 --> 00:00:39.149
from fixed assumptions, not measured prompts, tool responses, or provider billing.

00:00:39.149 --> 00:00:43.385
The calculator calls the final input of nine thousand its peak context.

00:00:43.385 --> 00:00:47.620
Its warning compares that input-only peak to the entered limit. Setting the

00:00:47.620 --> 00:00:51.856
limit to nine thousand produces no warning. Including the final four hundred

00:00:51.856 --> 00:00:56.092
output tokens would instead make nine thousand four hundred. At point zero

00:00:56.092 --> 00:01:00.328
three pixels per context token the bars span two hundred seventy and

00:01:00.328 --> 00:01:04.564
two hundred eighty two. The source does not include newly generated output

00:01:04.564 --> 00:01:08.800
in its limit check, and the warning does not stop calculation. A

00:01:08.800 --> 00:01:13.036
missing warning therefore cannot prove that a real model request fits a

00:01:13.036 --> 00:01:17.272
provider context window.

00:01:17.272 --> 00:01:21.577
With eight steps and a one thousand token history window, the first

00:01:21.577 --> 00:01:25.881
input is two thousand and the next seven are three thousand each.

00:01:25.881 --> 00:01:30.186
Total input falls from forty four thousand to twenty three thousand. At

00:01:30.186 --> 00:01:34.491
point zero zero six pixels per modeled input token the bars shrink

00:01:34.491 --> 00:01:38.796
from two hundred sixty four to one hundred thirty eight. The two

00:01:38.796 --> 00:01:43.100
thousand token system prompt remains outside the history cap, and outputs still

00:01:43.100 --> 00:01:47.405
total thirty two hundred. Costs multiply these totals by illustrative input and

00:01:47.405 --> 00:01:51.710
output prices, then by runs, experiments, and thirty days. Current provider pricing,

00:01:51.710 --> 00:01:56.015
cached tokens, execution failures, and memory quality are not verified by this

00:01:56.015 --> 00:02:00.319
calculation.
