Rates verified July 27, 2026
Claude Fable 5 token counter
Paste text, get an estimated token count and what it costs across the current Claude models. The estimate runs in your browser, which is why it is instant and why it is approximate.
What you get, in one paragraph
This page splits your text with a heuristic and multiplies the result by published rates — Fable 5 at $10 per million input tokens and $50 per million output, Opus 5 at $5 and $25, Sonnet 5 at its introductory $2 and $10, Haiku 4.5 at $1 and $5, all verified July 27, 2026. That is close enough to size a budget and not close enough to reconcile an invoice. The only exact count is Anthropic's /v1/messages/count_tokens endpoint, called with your real prompt and model ID.
Estimate a token count
Characters
0
Words
0
Estimated tokens
0
Estimated cost
$0.00
What this counter does not know
It does not know how the real tokenizer will split your text. It uses a character-and-word rule calibrated on English prose, so it drifts on code, JSON, XML and YAML, and it drifts on Chinese, Japanese and other non-Latin scripts by an amount we have not measured. It does not know your output length — that is set by the effort parameter, which moved one published prompt from about 1,900 output tokens at low to about 14,400 at max. It does not know whether a request will be answered by Fable 5 or handed to Opus 4.8 by a safety classifier, which changes what you are billed. Use it to size a budget, then confirm with count_tokens.
Turn a token count into a bill
A token count on its own says little. The effort setting, prompt caching at $1 per million read on Fable 5, and the 50% Batch discount all move the total more than the raw count does.