Token counter

Paste text and count LLM tokens in your browser. OpenAI uses the real o200k tokenizer; other providers show labeled estimates. Nothing you paste is uploaded.

Tokens

0

Characters

0

Words (approx.)

0

Pick a provider and paste text to count. OpenAI models use the real o200k tokenizer; other providers show labeled estimates.

Estimate API cost for this count →

How it works

  1. Choose the provider and model you will actually call — tokenizers differ.
  2. Paste a real prompt or sample output into the box above.
  3. Read the token count. For OpenAI it is exact (o200k). For Anthropic, Google and others it is an estimate and labeled as such.
  4. Need dollars, not just tokens? Open theAPI cost calculatorwith the same model, or jump to a model page below.

Count tokens, then check API cost

Official prices only, verified 2026-10-27. Rankings elsewhere on this site are sorted by price only.

GPT-6 Astra cost calculatorOpenAI · $10 in / $50 out per 1MGPT-6.1 Sol cost calculatorOpenAI · $2 in / $10 out per 1MClaude Sonnet 5.5 cost calculatorAnthropic · $2 in / $10 out per 1MClaude Haiku 5.5 cost calculatorAnthropic · $0.1 in / $0.5 out per 1MGemini 4 Argon cost calculatorGoogle · $2 in / $10 out per 1MGPT-6 Luna cost calculatorOpenAI · $0.1 in / $0.5 out per 1MClaude Opus 5.5 cost calculatorAnthropic · $4 in / $20 out per 1MGemini 3.1 Pro cost calculatorGoogle · $2 in / $12 out per 1M

Browse all models,providers, or thefull LLM API cost calculator.

Frequently asked questions

How does this token counter work?

You pick a provider and model, paste text, and we count tokens in your browser. OpenAI models use the real o200k_base tokenizer (via gpt-tokenizer). Anthropic, Google and other providers have no official client-side tokenizer for current models, so those counts are clearly labeled estimates based on character density.

Is my text uploaded anywhere?

No. All counting runs locally on your device. The page never sends your pasted text to our servers or any third party.

Why are some counts labeled “estimated”?

Only OpenAI publishes a tokenizer we can run fully in the browser for current models. For Anthropic and Google we use character-based estimates and mark them as such, so you are never shown a fake “exact” number.

How do I turn a token count into an API cost?

Use the link below the counter to open the CostMyTokens calculator with the same model pre-selected, or go to that model’s page. Official prices only — we do not invent rates.

What is the difference between a token and a word?

A token is a sub-word unit used by LLM APIs for billing and context limits. English prose is often roughly 0.75 words per token; code and JSON are denser. Always measure a real sample with the tokenizer of the provider you will call.