Skip to content
S/T
mediumneeds a human

This model counts your prompt differently

The same text is not the same number of tokens on every model. Upgrades can raise your bill silently.

A token is a unit of a specific tokenizer, not a unit of text. Two models can read an identical prompt and bill you for different amounts, because their vocabularies were built differently.

Anthropic documents that the tokenizer introduced with Claude Opus 4.7 produces approximately 30% more tokens for the same text than the previous Claude tokenizer. That is a real effect with a real invoice consequence: a team that moved from Sonnet 4.6 to a 4.7-or-later model without touching a character of their prompt saw their input token count rise by roughly a third.

The trap is that per-token prices are what get compared when people evaluate a migration. A model at the same headline price per million tokens is not the same price per request if it counts more tokens.

When you compare models, compare cost per request on your actual prompt — which is what the comparison table on this page does — rather than cost per million tokens.

Before
Sonnet 4.6: 4,000 tokens per request at $3/MTok.
After
A 4.7-or-later model: roughly 5,200 tokens for the same text. Same words, 30% more tokens.

Check your own prompt

The analyser checks this pattern along with the other 25, prices each finding against your request volume, and hands back a rewritten prompt. It runs in your browser — nothing is uploaded.

Run the analyser

More on model choice