AI & LLM
Token, cost & context tools for LLMs
GPT Token Counter — Paste Text, Get Exact Token Count
GPT models use byte-pair encoding (BPE) tokenization: words are split into subword pieces, so the token count for English text is typically 25–33% higher than the word count. For the sentence "The quick brown fox jumps over the lazy dog.", the GPT (o200k) count is 11 tokens ≈ 8 words. GPT-5.5 costs $5.00 per million input tokens and has a 1.1M-token context window; GPT-4o costs $2.50/M input with a 128K window. A 1,000-word essay is roughly 1,333 tokens. Paste your prompt or document to count before calling the API.
Sample: "The quick brown fox jumps over the lazy dog." → 11 tokens, 8 words
- 1,000-word essay ≈ 1,333 tokens (words ÷ 0.75)
- GPT-5.5 context window: 1,050,000 tokens
- GPT-5.5 input price: $5.00/M tokens
- Tokenizer: O200K (exact in-browser count for OpenAI models)
Frequently asked questions
How many tokens per word in GPT models?
On average, 1 word ≈ 1.33 tokens in GPT (o200k/cl100k). Equivalently, 1 token ≈ 0.75 words. Common English words ("the", "a", "is") are usually single tokens; longer or technical words split into 2–4 pieces. Code and non-English text typically use more tokens per word than English prose.
Does GPT-5.5 use the same tokenizer as GPT-4o?
Yes — both GPT-5.5 and GPT-4o use the o200k_base vocabulary (200,000 token types). Earlier models (GPT-4-turbo, GPT-3.5-turbo) used cl100k_base (100,000 types). The o200k tokenizer is more efficient on code and non-English text. This calculator selects the correct encoding when you pick a model from the dropdown.