PastelyArchlyv2.3
Timecard
Get TimecardGet $5
Free toolNo sign-up · Runs in your browser

AI token counter

A token is a chunk of text a model reads, often part of a word. In English, one token is about 4 characters, and 100 tokens is about 75 words. Paste your text to count it exactly with OpenAI's o200k tokenizer and see what it costs on today's models.

Tokens

≈38

Characters

150

Words

26

HOW THE MODEL SEES IT · o200k_base

Loading the tokenizer…

COST AS INPUT

  • GPT-5.5$5 in · $30 out per 1M$0.00019
  • GPT-5.6 Terra$2 in · $12 out per 1M$0.000076
  • GPT-5 mini$0.25 in · $2 out per 1M$0.0000095
  • Claude Opus 5.5$4 in · $20 out per 1M≈$0.00015
  • Claude Sonnet 5.5$2 in · $10 out per 1M≈$0.000076
  • Claude Haiku 5.5$0.1 in · $0.5 out per 1M≈$0.0000038

GPT counts are exact. Claude prices use the same count as an estimate, since its tokenizer isn't public. List prices, October 2026.

How many tokens is a word?

In English, a token is about 4 characters, or three quarters of a word: 100 tokens is roughly 75 words, and a 1,500-word article is about 2,000 tokens. Code, numbers and other languages use more tokens per word.

TextTokens (o200k)
hello1
Timecard2
unbelievably3
Hello, world!4
const total = hours * rate;7
Où est la gare ?6

Source: OpenAI, “What are tokens and how to count them?”, and the o200k_base tokenizer.

AI model prices per million tokens

List prices for input and output, the same ones Timecard uses to work out the API value of your AI coding sessions.

ModelInputOutput
GPT-5.5$5$30
GPT-5.6 Terra$2$12
GPT-5 mini$0.25$2
Claude Opus 5.5$4$20
Claude Sonnet 5.5$2$10
Claude Haiku 5.5$0.1$0.5

Why your token count matters

Every message in a conversation is sent again with the next one, so long chats and big files multiply the token count fast. Coding agents like Claude Code and Codex can read millions of tokens in a day. Timecard reads their history on your Mac and shows tokens, time and API value per session and per project.

AI Token Counter
questions

What is a token?

Models read text in tokens, not words. Common words are one token, longer or rarer words split into several, and spaces and punctuation count too. OpenAI's rule of thumb for English is one token per 4 characters, or about 75 words per 100 tokens.

Which tokenizer does this use?

o200k_base, the tokenizer OpenAI uses for GPT-4o, the o-series and GPT-5 models. Counts for those models are exact. It runs in your browser, so your text never leaves the page.

Are Claude token counts exact?

No. Anthropic hasn't published Claude's tokenizer, so Claude counts here are an estimate from the same text. Claude usually counts somewhat more tokens than o200k for the same English. For an exact count, use Anthropic's token counting API.

How is the cost worked out?

Tokens × the model's list price per million tokens, as input. Output is billed separately and usually costs several times more per token. Prices are list prices as of October 2026 and don't include caching discounts.

How do I see what my AI coding actually costs?

Timecard reads the history Claude Code, Codex, Cursor, opencode and Gemini CLI keep on your Mac and adds up tokens and API value for every session, without uploading anything.

More free tools

Claude Usage Reset Timer

When your Claude or Codex limit resets, with a live countdown.

Open →

Claude Code Cost Calculator

What Claude Code would cost on the API each month, next to Pro and Max.

Open →

Screen Time Calculator

How many years of your life will you spend on screens?

Open →

Time Card Calculator

Add up a week of hours, breaks, overtime and pay.

Open →

Military Time Converter

Convert 24-hour military time to standard time, with a full chart.

Open →

Hours Between Two Times

Find the hours and minutes between any two times, even past midnight.

Open →

Work Hours in a Year

How many working hours and days are in a year, after holidays and time off.

Open →

Decimal Hours Calculator

Convert hours and minutes to decimal hours, and back.

Open →

Time Blocking Planner

Plan your day in blocks, see where the hours go, and print it.

Open →