ChatGPT Token Cost
Paste text to estimate its token count and the API cost on GPT models. A quick heuristic, not an exact tokenizer.
Formula last reviewed 4 August 2026 · How we verify our calculators
Estimated input tokens
- Estimated cost (USD)
- $0.00754
- Input cost
- $0.00003
- Output cost
- $0.00750
Updates live as you type
Frequently asked questions
It is a rough heuristic that blends character and word counts. Real token counts depend on the model’s actual tokenizer, so treat this as a ballpark — usually within ~20%.
A token is a chunk of text the model processes — roughly 4 characters or 0.75 words of English on average. Pricing is per million tokens, split into input and output rates.
Output (generated) tokens cost more than input (prompt) tokens on most models, so the calculator prices them separately. Enter how long you expect the reply to be.
This one starts from pasted text to estimate tokens; the OpenAI cost calculator starts from token volumes to project monthly spend. Use this for one-off checks.
Fourteen tokens in, and the reply is what actually costs you
Paste "Write a short product description for a calculator website." and the heuristic here estimates roughly 14 input tokens. Pair that with 500 expected output tokens on GPT-5.4 (priced at $2.50/M input, $15/M output), and the input cost is a fraction of a cent while output cost dominates the total — 500 output tokens at $15/M outweighs 14 input tokens at $2.50/M by a wide margin. That pattern holds broadly: for short prompts paired with longer replies, output pricing drives the bill, not how long your prompt was.
Why this is a heuristic, not an exact tokenizer
Token estimation here blends character and word counts to approximate what a real tokenizer would return — a useful ballpark, typically within about 20% for ordinary English text, but not a substitute for the model's actual tokenizer, which is the only source of an exact count. Choose a model, set expected output tokens, and see input, output and total cost broken out separately, since pricing is per million tokens with output usually costing more than input on every model.
Where this tool fits versus the others
This is built for one-off, paste-and-check estimates — pasting a real prompt before sending it to the API gives a cost sense without spending an actual API call to find out. For long, repeated prompts, prompt caching and batch processing can cut costs substantially, covered in more depth on the OpenAI cost calculator. To project ongoing spend across many requests over a month rather than checking a single prompt, the OpenAI cost calculator works from request volume and average token sizes instead — the better tool once you're past a one-off check.
Pricing as of June 2026. LLM rates change frequently — verify current prices on the provider's official pricing page before budgeting.