Start here · The whole machine
One answer takes many trips
Your message enters as text. The tokenizer cuts it into tokens, then the rest of the model works with those tokens. At the far end, the model picks one new token and adds it to the message.
The machine runs again to pick the next token, then again for the token after that. The input meter counts what the model reads. The output meter counts each new token it writes.
FIGURE 0 · ONE TRIP THROUGH THE MACHINEautoplays
loading…
TEXT IN → CUT INTO TOKENS → MODEL → ONE TOKEN OUT → REPEAT · COST LINK: INPUT COUNTS WHAT THE MODEL READS; OUTPUT COUNTS EACH TOKEN IT WRITES
Activity 1 · See the meter
Put a sentence through the tokenizer
A token can be a whole word, part of a word, a number, or punctuation. Common text often fits into larger pieces. Less familiar text often breaks into more pieces. More pieces mean more tokens on the bill.
1
What to do
about 1 minute
Pick a sample. Look at the blue pieces. Then type your own sentence and watch the token count change.
ACTIVITY 1 · SEE THE CUTtype or pick a sample
One blue piece = one token
Words0
Tokens0
Extra pieces0
Example rate · $5 per million input tokens · $25 per million output tokens
To send this text$0
To write this much$0
THIS PAGE USES A SMALL ENGLISH-HEAVY TOY TOKENIZER · COST LINK: TOKEN COUNT × NUMBER OF USES × RATE
Activity 2 · Test your instinct
Which message costs more?
You do not need to guess an exact token count. Pick the message that you think will use more tokens. The reveal shows both cuts and the monthly cost if a service sends that message one million times.
2
What to do
about 2 minutes
Read A and B. Pick the one that you think costs more. There are three rounds.
ACTIVITY 2 · PICK THE COSTLY ONEround 1 of 3
Which costs more to send?
NO EXACT GUESS REQUIRED · EXAMPLE USE: 1 MILLION SENDS PER MONTH AT $5 PER MILLION INPUT TOKENS · COST LINK: SMALL DIFFERENCES REPEAT
Activity 3 · Change the bill
Get the monthly bill below $55,000
This help-desk assistant handles 100,000 requests a day. Each request includes repeated instructions, background material, and an answer. Start at $82,500 a month, then choose any two changes.
3
What to do
about 2 minutes
Turn on two changes. Watch which part of the bill shrinks and whether you reach the target.
ACTIVITY 3 · LOWER THE BILLtarget: under $55,000
100,000 requests a day · $5 per million input tokens · $25 per million output tokens
$82,500per month· $0 saved
repeated instructions $22,500
background $30,000
answers $30,000
Choose two changes to reach the target.
MONTHLY COST = INPUT COST + OUTPUT COST · COST LINK: OUTPUT TOKENS USE THE HIGHER EXAMPLE RATE, SO SHORTER ANSWERS CAN SAVE MORE THAN EXPECTED
Next · The transformer videos
Follow the middle of the machine
The videos start where Activity 1 ends. Tokens become vectors, the transformer mixes information across the message, and the model produces scores for the next token. When the model picks one, return to the loop above: one output token has been billed, and the machine goes around again.
Leave with this
Three facts explain the token bill
- Tokens are pieces of text. They do not line up neatly with words.
- The bill has two meters. Input is what the model reads; output is what it writes.
- Small counts become large costs through repetition. Text sent on every request deserves the closest look.
About this page: The tokenizer is a small English-heavy teaching model. Production tokenizers have much larger vocabularies, and each provider uses its own rates. The dollar figures here use one fixed example rate so the exercises stay easy to compare.