Token Counter for AI Prompts
Paste a prompt, get an honest token estimate — a range, not a fake-precise number — plus the cost across GPT, Claude, and Gemini.
Overview
A token is the unit models actually read and bill, and it is neither a character nor a word. This loads a typical assistant prompt and counts it: an estimated token range (because tokenizers genuinely disagree on the same text), the same text's count across four models, and what one call costs versus a thousand. The number is honest about its own uncertainty — that is the point of an estimator that refuses to lie with false precision.
Workflow
-
Paste the prompt
Any text — the report shows counts only, never echoes your content.
-
Read the range
Tokenizers disagree, so the estimate is a range with the model comparison beside it.
-
Check the cost
Per call looks trivial; the per-1,000-calls line is the real budget.
Why This Works
- A range is honest where a single number would be a lie — tokenizers really do differ
- Content-type detection adjusts the estimate for prose, code, and CJK
- The cost line turns an abstract count into a number you can budget against
Best for
- Getting a quick, honest token count
- Sanity-checking a prompt before an API call
- Comparing the same text across models
Not for
- Deciding whether content fits a context window — that's the Context Window Estimator
- Counting characters or words for a text limit — that's the Character Counter
Use cases
- Getting a quick, honest token count
- Sanity-checking a prompt before an API call
- Comparing the same text across models