Cheapest output
Claude Haiku 5.5
Input $0.1 per 1M · Output $0.5 per 1M
To read it (input)$0.10
To write it (output)$0.50
AI Price Watch
Tokens, translated
AI is priced in tokens, a unit nobody can picture. This page translates tokens into the units everybody already knows: words, pages, novels, hours of reading, weeks of typing, and dollars.
The scale explorer
Drag the slider. Every number below is the same amount of tokens, translated six ways, then priced three ways.
1,000,000 tokens
1,0001 billion
Try:
750,000Words
1,500Pages
8.3Novels
50 hoursReading it
313 hoursTyping it
7.8Typing, in work weeks
What that amount costs
Three reference models from the current price list, picked by output price: the cheapest, the middle of the list, and the most expensive.
Mid priced
GPT-6.1 Terra Pro
Input $1 per 1M · Output $5 per 1M
To read it (input)$1.00
To write it (output)$5.00
Most expensive
Claude Fable 5.1
Input $10 per 1M · Output $50 per 1M
To read it (input)$10.00
To write it (output)$50.00
What tokens actually do
Rough budgets. Real projects vary, sometimes a lot. Tap a card to load its budget into the explorer and see it translated.
The scale map
The same slider, drawn. Watch a handful of tokens grow into words, pages, novels, shelves, a whole library, then a street of libraries.
- Tokens
- Words
- Pages
- Novels
- Shelves
- A library
- A street of libraries
Novels: about 8.3 novels standing on one shelf.
The price landscape
Every model on the current list, ranked by output price: what one million written tokens cost. Output is the expensive direction, so it sets the shape of the landscape. Tap a model to price the explorer amount on it.
Output price per 1M tokens, log scale
Pick a model above to see its input price and what the explorer amount would cost on it.
Tokens, translated: the table
The whole ladder on one screen, for the readers who like their translations in rows.
| Tokens | Words | Pages | Novels | Reading | Typing | Output cost, cheapest (Claude Haiku 5.5) | Output cost, priciest (Claude Fable 5.1) |
|---|---|---|---|---|---|---|---|
| 1,000 | 750 | 1.5 | 0.01 | 3 min | 19 min | $0.0005 | $0.05 |
| 10,000 | 7,500 | 15 | 0.08 | 30 min | 3.1 hours | $0.0050 | $0.50 |
| 100,000 | 75,000 | 150 | 0.83 | 5 hours | 31 hours | $0.05 | $5.00 |
| 1,000,000 | 750,000 | 1,500 | 8.3 | 50 hours | 313 hours | $0.50 | $50.00 |
| 10,000,000 | 7,500,000 | 15,000 | 83 | 500 hours | 3,125 hours | $5.00 | $500.00 |
| 100,000,000 | 75,000,000 | 150,000 | 833 | 5,000 hours | 31,250 hours | $50.00 | $5,000.00 |
| 1,000,000,000 | 750,000,000 | 1,500,000 | 8,333 | 50,000 hours | 312,500 hours | $500.00 | $50,000.00 |
The math on this page: 1 token is about 0.75 words; a page holds about 500 words; a novel runs about 90,000 words; reading speed 250 words a minute; typing speed 40 words a minute; a work week is 40 hours. Approximations, stated once and used everywhere.
Provider list prices, as of November 2, 2026. Prices move; the translations do not. Educational only, not financial advice.
Input and output, once more
Input tokens are what you hand the model to read: your question, your documents, the conversation so far. Output tokens are what it writes back. Writing takes more work than reading, which is why the output column costs more than the input column on every model in the landscape above, with no exceptions in the current list.
That one fact explains most AI pricing you will ever meet. Reading a whole novel costs a model pennies. Writing one is where the money goes, and it is still, as the explorer shows, a fraction of what the typing would cost you in time.
Live prices
Trends
The price table itself, with the watchlist charts, dated and sourced, refreshed weekly.
See the price table →Your own usage
The Tools calculator
Type in your own monthly tokens and get the bill per model, cheapest first, in dollars.
Open the calculator →