LLM Toolkit Token ↔ Words Converter
1,000 tokens ≈ 750 words in English — that’s the headline. Enter any amount below for a precise conversion, or scan the reference table for common sizes.
Result
—
English prose at 0.75 words per token. Other languages vary — see FAQ below.
Token to words reference table
| Tokens | ≈ Words | ≈ Characters | ≈ Pages | Fits in |
|---|---|---|---|---|
| 100 | 75 | 400 | 0.1 | every current model |
| 250 | 188 | 1,000 | 0.4 | every current model |
| 500 | 375 | 2,000 | 0.8 | every current model |
| 1,000 | 750 | 4,000 | 1.5 | every current model |
| 2,000 | 1,500 | 8,000 | 3.0 | every current model |
| 4,000 | 3,000 | 16,000 | 6.0 | every current model |
| 8,000 | 6,000 | 32,000 | 12 | every current model |
| 16,000 | 12,000 | 64,000 | 24 | every current model |
| 32,000 | 24,000 | 128,000 | 48 | every current model |
| 64,000 | 48,000 | 256,000 | 96 | every current model |
| 128,000 | 96,000 | 512,000 | 192 | Claude / GPT-5 / Gemini (200K+) |
| 200,000 | 150,000 | 800,000 | 300 | GPT-5 / Gemini (400K+) |
| 400,000 | 300,000 | 1,600,000 | 600 | Gemini only (1M) |
| 1,000,000 | 750,000 | 4,000,000 | 1500 | exceeds all current windows |
Estimates assume English prose: 0.75 words and 4 characters per token, 500 words per single-spaced page. The "fits in" column shows the smallest current model window that holds this many tokens with room for a response.
Frequently asked questions
How many words is 1,000 tokens?
Approximately 750 words of English text. English averages 1.33 tokens per word (0.75 words per token), so 1,000 tokens ≈ 750 words ≈ 4,000 characters ≈ 1.5 pages. The ratio holds well for prose but not for code, which tokenizes denser.
How many tokens is 500 words?
About 665 tokens. At 1.33 tokens per English word, multiply your word count by 1.33 to estimate tokens. A typical blog post of 1,500 words runs ≈ 2,000 tokens; a 60,000-word manuscript runs ≈ 80,000 tokens — which fits comfortably in every current model’s context window.
How many tokens is a page of text?
A single-spaced page of English (≈500 words) is roughly 665 tokens. Double-spaced academic pages (≈250 words) run ≈330 tokens. So a 30-page document is about 20,000 tokens — small enough that most models can hold several documents at once.
Do different models have different words-per-token ratios?
Slightly, yes. Modern tokenizers (GPT-5’s o200k, Claude’s, Gemini’s) are all BPE-family and converge near 0.75 words per token for English. The big differences appear in other languages: Spanish and French run 1.5–2 tokens per word, German slightly higher, and CJK languages can exceed one token per character — meaning 1,000 tokens buys you far less Chinese text than English.
How many characters per token?
About 4 characters per token for English, including the space after each word. Common words ("the", "and", "is") are single tokens; rarer words split into 2–4 tokens. Numbers and unusual Unicode sequences tokenize less efficiently.
Why does this matter for API costs?
Because you pay per token, not per word. Converting your content budget to tokens tells you your real API spend: 100 articles of 1,500 words each ≈ 200,000 tokens ≈ $0.25 to process on GPT-5 mini, but $2.50 on GPT-5 and $0.01 on GPT-5 nano. Measure your actual text with our Token Counter for exact numbers.
Measure, don’t estimate
Ratios get you close; the AI Token Counter gets you exact — it runs OpenAI’s real tokenizer in your browser. Then price the result with the API Cost Calculator.