// tokenizer

LLM token counter

Paste a prompt, a document or a model's answer to count its tokens, then see what that text costs to send or generate on 336 models at OpenRouter list prices.

Tokens (o200k_base)
Characters
244
Words
44
Characters per token

What this text costs

The count above is exact for OpenAI's o200k_base tokenizer. Other providers tokenize differently — Anthropic and Google don't publish theirs — so their rows (marked ≈) apply this count as an estimate; expect their real counts to differ by roughly 10–30%.

Cost of the text as input or output, per call and for the chosen number of calls
ModelInput / 1MAs input, 1 callAs input, 1,000 callsAs output, 1,000 calls
Claude Fable 5.1anthropic$10.00
Claude Opus 5.5anthropic$4.00
Claude Fable 5anthropic$10.00
GPT-6 Astraopenai$10.00
Muse Spark 1.3meta$1.25
DeepSeek V4.1 Flashdeepseek$0.100
GPT-5.6 Solopenai$2.00
GPT-5.5openai$5.00
Claude Opus 5anthropic$5.00
GPT-6 Solopenai$2.00
Kimi K3moonshotai$3.00
Gemini 3.7 Flashgoogle$0.750

Token counter FAQ

What is a token?

Models read and write text as tokens — chunks produced by a tokenizer, usually a common word, part of a longer word, or a piece of punctuation. In English, one token averages about four characters, or roughly three quarters of a word, but code, non-English text and unusual formatting use noticeably more tokens per character.

Which tokenizer does this counter use?

OpenAI’s o200k_base encoding, run in your browser with js-tiktoken. The count is exact for OpenAI models that use o200k_base. Anthropic and Google do not publish their tokenizers, so for their models the count is an estimate — their real counts for the same text commonly differ by 10–30%, and the cost table marks those rows with ≈.

Is my text sent anywhere?

No. Tokenization runs entirely in your browser after the tokenizer’s vocabulary file loads. Nothing you paste is uploaded, logged or stored.

Why do output tokens cost more than input tokens?

Input is processed in one parallel pass, while each output token needs its own forward pass through the model. Providers price accordingly — output typically costs three to six times more than input — and reasoning tokens a model generates internally bill as output too.