1 million tokens to words
1 million tokens is approximately 750,000 English words — about 4,000,000 characters, or 1,417–1,625 single-spaced pages. The exact count depends on the tokenizer and your text: code, heavy punctuation, and non-English content use more tokens per word.
The 1 million-token breakdown
| Measure | Range | Typical |
|---|---|---|
| Words | 650,000–850,000 | 750,000 |
| Characters | 3,000,000–5,000,000 | 4,000,000 |
| Single-spaced pages | 1,417–1,625 | 1500.0 |
Ranges reflect the 0.65–0.85 words-per-token spread between casual and dense technical English prose.
What 1 million tokens means for context
- Claude Sonnet 5 (Anthropic, 1,000,000 window): 100% used
- Claude Haiku 4.5 (Anthropic, 200,000 window): exceeds the window
- GPT-5.6 Sol (OpenAI, 1,050,000 window): 95% used
- GPT-5.6 Luna (OpenAI, 1,050,000 window): 95% used
- Gemini 3.1 Pro Preview (Google, 1,048,576 window): 95% used
- Grok 4.6 (xAI, 500,000 window): exceeds the window
- Kimi K3 (Moonshot, 1,048,576 window): 95% used
- DeepSeek V4 Flash (DeepSeek, 1,048,576 window): 95% used
Context is shared with the model’s reply — a prompt that fills the window leaves no room for the answer.
Make it exact
Ratios are planning tools. When the number matters, count your actual text: the prompt calculator runs a real BPE tokenizer in your browser — exact for OpenAI models, a labeled approximation for Claude, Gemini, Grok, Kimi, and DeepSeek.
1 million tokens, converted
How many words is 1 million tokens?
1 million tokens (1,000,000) is roughly 750,000 English words, with a realistic planning range of 650,000–850,000 words depending on writing style, vocabulary, and language.
How many characters is 1 million tokens?
About 4,000,000 characters including spaces (the standard 4-characters-per-token rule). Code and non-Latin scripts usually need more characters per token, so their counts run higher.
How many pages is 1 million tokens?
Roughly 1,417–1,625 single-spaced pages at 400–600 words per page, based on the 750,000-word equivalent.
Why do different tools give different word counts for the same tokens?
Token counts depend on each model family’s tokenizer vocabulary, and word-count conversions are ratios applied on top. A range with a typical value — like the table above — is the honest representation.
Free token tools. Everything computes in your browser.