← Back to counter

How many tokens is a word?

The short answer

In typical English text, 1 word ≈ 1.3 tokens across most modern AI tokenizers (GPT-4o, Claude, Gemini). Or, conversely, 1,000 tokens ≈ 750 words.

Single short common words ("the", "and", "cat") are usually 1 token. Longer or less common words ("antidisestablishmentarianism", "tokenization") split into 2-5 pieces.

Why words don't map cleanly to tokens

Tokenizers don't split on whitespace — they split on subword units learned from training data. The most common letter sequences become single tokens; rare sequences split into multiple.

A few examples (GPT-4o's o200k_base tokenizer):

• cat → 1 token
• tokenizer → 1 token (common in AI training data)
• tokenization → 2 tokens (token + ization)
• pneumonoultramicroscopicsilicovolcanoconiosis → 13 tokens

Punctuation, leading whitespace, and capitalization all affect tokenization too. cat (with leading space) is a different token than cat.

Practical estimation

Rough rules of thumb for English:

For non-English languages, the ratio shifts substantially:

• Chinese / Japanese with o200k_base tokenizer: ~2-3 tokens per character (vs. ~1.3 per English word)
• Code: 10-20% more tokens than equivalent prose
• Numbers and identifiers: usually 1 token per ~3 digits

Get an exact count

Paste your text into the counter — it shows the exact token count for your input across every supported model. Different models have different tokenizers, so the count varies by 5-20% across providers for the same text.

Try this on every model

• Claude Opus 4.7 $15.00/$75.00
• Claude Sonnet 4.6 $3.00/$15.00
• Claude Haiku 4.5 $0.80/$4.00
• GPT-4o $2.50/$10.00
• GPT-4o mini $0.15/$0.60
• GPT-4 Turbo $10.00/$30.00
• Gemini 2.5 Pro $1.25/$10.00
• Gemini 2.5 Flash $0.07/$0.30
• Llama 3.1 405B $3.50/$3.50
• Llama 3.1 70B $0.59/$0.79
• Llama 3.1 8B $0.18/$0.18
• Mistral Large $2.00/$6.00
• DeepSeek V3 $0.27/$1.10
• Qwen 2.5 72B $0.90/$0.90
• Qwen 2.5 Coder 32B $0.80/$0.80

Try the live counter →