Estimator
Token Counter
Tokens decide what fits, what it costs and what gets truncated. This estimates the count for any text you paste, prices it across the common models, and tells you whether it fits the context window you are working with.
- Counts and pricesBoth, in one place
- Context fit shownWhat gets truncated
- Nothing transmittedCounted in your browser
Check it
Paste it here
Result
- Words —
- Characters —
- Fits the window —
- Share of window used —
- Cost per call, budget model —
- Per month at this volume —
Token counts are estimated at roughly four characters per token for English prose, which is close for ordinary text and drifts on code, markup and other languages. For billing precision, use your provider's own tokenizer.
What the result means
Your text is not the only thing in the window
The system prompt, the conversation so far and every retrieved chunk share the same budget. A document that fits alone may not fit in the request that actually runs.
Four characters per token is a rule of thumb
It holds well for English prose and drifts on code, JSON, markup and non-Latin scripts, which pack more tokens per character. Treat this as a planning number.
A big window is not a reason to fill it
Cost scales with what you send, and retrieval quality usually falls as you add marginal context. Fitting is the constraint; filling is a choice, and often a bad one.
Truncation happens silently
Most systems drop the middle or the oldest turns without telling anyone. If answers degrade on long conversations, this is usually why.
Questions
Close for English prose, within a few percent. It drifts on code and other languages. For billing precision use your provider's tokenizer.
Every model family uses its own tokenizer. The counts are similar but not identical, which is why this is labelled an estimate.
No. It is counted in your browser and nothing is transmitted.
Want this run against your whole site?
Tell us the page your assistant lives on and the questions you care about. A person runs them and sends back the raw answers.