Token Counter

Estimate AI tokens and context usage.

Paste text to estimate token count, context usage, and approximate cost for common AI model profiles.

EstimateReady
0input tokens
0%context used
0tokens left
$0.00estimated cost
Paste text to estimate token usage.

How it works

The estimator combines character count, word count, line count, whitespace density, and your selected profile to approximate token usage and context fit locally.

Token counts are estimates. Use provider usage data or an exact tokenizer for billing-critical decisions.

Practical examples

  • If a context limit is 128,000 tokens, an estimated 20,000-token input plus a planned 4,000-token answer leaves 104,000 tokens.
  • Compare the general-text and code profiles for the same pasted snippet to see how text structure changes the estimate.
  • Set expected output to zero when you only want a rough input-size check.

How to read the result

  • Input tokens are an estimate, while context used compares that estimate with the full selected context limit.
  • Tokens left subtracts both estimated input and planned output, so a negative value means the plan exceeds the chosen context.
  • Estimated cost applies the selected planning rates per one million input and output tokens.

Limits and privacy notes

  • Tokenization depends on the exact model and tokenizer; use provider usage data for billing or hard limits.
  • The economy, standard, and premium prices are editable planning profiles, not current provider quotes.
  • Prompt text is analyzed in the page and is not uploaded.

Method used

  • General text uses characters ÷ 3.9; code uses characters ÷ 3.1; compact prompts use characters ÷ 4.4.
  • The estimate also adds 0.08 token per whitespace-delimited word and 0.35 token per non-empty line, then rounds up.
  • Remaining context = context limit − estimated input − planned output.
Context fit

Plan input plus output

  • Remaining context subtracts both the prompt and expected response.
  • Increase expected output tokens when planning summaries or rewrites.
  • Use a provider tokenizer for billing-critical work.
Profiles

Pick the closest text type

  • Code and structured text usually tokenize denser than plain prose.
  • Compact prompts can estimate differently than transcripts.
  • Pricing profiles are planning references, not live prices.
Cost hygiene

Reduce before sending

  • Trim logs, duplicated context, and boilerplate first.
  • Check line-heavy inputs separately.
  • Copy the summary into planning notes.

Frequently asked questions

Why can a provider report a different token count?

Production tokenizers split text according to a model-specific vocabulary. This page uses a transparent approximation instead of claiming exact model parity.

What does a negative tokens-left result mean?

The estimated input plus planned output is larger than the selected context limit. Reduce either value or choose the correct context limit.

Are the price profiles current API prices?

No. They are fixed planning scenarios. Confirm current prices with the provider before making a budget decision.

Is pasted prompt text sent anywhere?

No. The estimate is calculated by the JavaScript in this page.

Copied