LLM Token Counter (OpenAI, Claude, Gemini)

What is Token Counter?

Large language models bill and limit context by tokens, not words — a token is roughly ¾ of a word in English, but code and non-English text tokenize differently. This counter estimates how many tokens your text uses across major model families and projects the API cost, so you can stay under context windows and predict spend before sending large prompts.

How to count LLM tokens

1. 1Paste your prompt or document into the input area.
2. 2See estimated token counts for GPT, Claude, and Gemini side by side.
3. 3Enter expected output length to project total cost.
4. 4Trim or split the input if it exceeds your model's context window.

Use Cases

Code Examples

~4 characters   ≈ 1 token
~0.75 words     ≈ 1 token
1,000 tokens    ≈ 750 words

Key Concepts

Token

A sub-word unit produced by the model's tokenizer. Common words are one token; rare words, code, and punctuation split into several.

Tokenizer differences

GPT-4/3.5 use cl100k_base (BPE); Claude uses a similar BPE vocabulary; Gemini uses SentencePiece. The same text yields slightly different counts (~10–15%).

Context window

The maximum tokens (prompt + completion) a model can handle in one call. Exceeding it truncates or errors.

Tips & Best Practices

• ▸Code and non-English text tokenize less efficiently than English prose — expect more tokens per character.
• ▸For billing-critical counts, use the official tokenizer (tiktoken for OpenAI) — estimates are within a few percent, not exact.
• ▸Both input AND output tokens are billed; account for the completion length in cost projections.
• ▸Whitespace and repeated punctuation consume tokens too — trimming boilerplate reduces cost.
• ▸To see which part of a long prompt uses the most tokens, use the prompt token breakdown tool.