Token count & cost estimator
Paste any prompt or text to estimate its token count and what it costs to run across models. A fast estimate — use your provider's tokenizer for exact counts.
Primary estimate uses ~4 characters per token; word-based ~range shown for comparison. This is an estimate (English ≈ 4 chars/token; code & other languages differ) — use the provider tokenizer for exact billing.
Model & run volume
Defaults = Claude Opus 4.8 published rates. Prices editable — verify at your provider.
Per-call cost comparison
| Model | $/1M in | $/1M out | Per call | Monthly |
|---|
Comparison uses each model's pre-filled price with your estimated input tokens, expected output and run volume. Edit a price and it recalculates.
From token count to LLM cost
Two questions come up constantly when you build with LLMs: how many tokens is this prompt, and what does it cost to run? This token counter answers the first with a fast heuristic — roughly four characters per token in English — and this prompt cost estimator answers the second by pricing your estimated input tokens plus your expected output tokens against each model's rate. It works as a GPT token calculator and a Claude token count tool alike, because the arithmetic is the same; only the price per million tokens changes.
Output tokens usually cost several times more than input, so a long answer can dominate a call's cost even when the prompt is short. That's why a good LLM token estimate looks at both sides. For exact billing always reach for the provider's official tokenizer — this is a planning estimate, not an invoice.
| Rule of thumb (English) | ≈ Tokens |
|---|---|
| 1 token | ~4 characters / ~0.75 words |
| 1 sentence | ~15–25 tokens |
| 1 paragraph | ~100 tokens |
| 1 page | ~500 tokens |
Rules of thumb for English prose. Code, non-English text and special characters tokenize differently — confirm with the provider tokenizer.
Counting tokens is the easy part.
Cutting them without losing quality is the work — prompt trimming, caching, output caps and model routing typically save 30–90%. Book a free build audit and we'll find where your token spend actually goes.
Book a Build AuditQuestions about tokens & cost
How many tokens is my text?
As a rough rule, one token is about four characters or three-quarters of a word in English, so a 1,000-character prompt is roughly 250 tokens. Code, non-English text and special characters tokenize differently, so use your provider's official tokenizer for exact billing — this tool gives a fast estimate.
How much does a prompt cost to run?
Multiply your estimated input tokens by the model's input price and your expected output tokens by its output price (both priced per million tokens), then multiply by how many times you run it. Output tokens usually cost several times more than input, so long responses dominate cost.
How do I reduce token costs?
Trim system prompts and context, summarize conversation history instead of resending it verbatim, cap max output tokens, cache stable prompt prefixes, and route easy requests to cheaper models. Together these commonly cut token spend 30–90% without lowering answer quality.