What counts toward a context window?

The model receives the system prompt, document, conversation and tool content, then generates output. This estimate adds the system prompt, document estimate, and expected output.

Are pasted-text token counts exact?

No. The one-token-per-four-characters estimate is a planning shortcut. Token boundaries differ by model, language, code, and formatting.

What does the chunking suggestion mean?

It divides the document estimate into pieces sized to leave room for the system prompt and expected answer. Review the actual request shape before sending.

Will this document fit in a model context window?

Estimate model context use before sending a document and planned output.

Example: Acme Inc: Paste a 1,200-word product review, allow 1,500 system-prompt tokens and a 1,000-token answer, then compare model context use.

Estimate the request size

Paste text or enter a document size. Your text stays in this browser.

Compare model context limits and API rates.

Document text

Example is ready. Counting uses one token per four characters.

Or enter a document size

System prompt tokensExpected output tokens

A page is estimated as 500 words. A word is estimated as 0.75 tokens. These are rough planning values, not tokenizer results.

Context window estimates for selected models
Model	Estimated tokens used	Window used	Fits?	Chunking suggestion
Anthropic · Claude Opus 5.5Provider model sourceChecked September 28, 2026	2,565 of 10,00,000	0.3%	Fits by this estimate	No split suggested
DeepSeek · DeepSeek V4.1 FlashProvider model sourceChecked September 28, 2026	2,565 of 10,00,000	0.3%	Fits by this estimate	No split suggested
Google · Gemini 3.1 Pro PreviewProvider model sourceChecked September 28, 2026	2,565 of 10,48,576	0.2%	Fits by this estimate	No split suggested

Chunking reserves 20% of each context window for request overhead. The estimate does not inspect or divide your text. Model data from models.dev (MIT).

Embed this tool

Copy this snippet to show the tool on your site.

How it works

Pasted text is estimated at one token per four characters. For a numeric size, one word is estimated at 0.75 tokens and one page at 375 tokens using 500 words per page.

Total estimated context is document tokens plus system prompt tokens plus expected output tokens. Percent used is that total divided by the published context window.

Chunk size uses 80% of the context window minus the system prompt and expected output to leave room for request overhead.

Sources and checked dates

OpenAI guide to counting tokensChecked 28 September 2026
Anthropic token countingChecked 28 September 2026
DeepSeek token usage notesChecked 28 September 2026
Mistral context limitsChecked 28 September 2026
Google Gemini model limitsChecked 28 September 2026
Qwen 3.8 Flash context and output limitsChecked 28 September 2026
Meta Llama context informationChecked 28 September 2026

Limits

Pasted text uses a rough one-token-per-four-characters estimate. Model tokenizers count the same text differently.
The estimate omits chat wrappers, tools, images, audio, retrieval additions, and hidden provider instructions.
Provider limits and available output space can change. Check the linked model page before sending a large request.

Common questions

What counts toward a context window?: The model receives the system prompt, document, conversation and tool content, then generates output. This estimate adds the system prompt, document estimate, and expected output.
Are pasted-text token counts exact?: No. The one-token-per-four-characters estimate is a planning shortcut. Token boundaries differ by model, language, code, and formatting.
What does the chunking suggestion mean?: It divides the document estimate into pieces sized to leave room for the system prompt and expected answer. Review the actual request shape before sending.

AI token counter
Count tokens in pasted text and estimate the input cost for a selected model.
AI API cost calculator
Estimate API token spend for several models using published input and output rates.
AI task brief generator
Structure a rough request into a brief a person or AI operator can review.

OperatorNest can take on repeat work, check with you before consequential steps, and leave a receipt for the result. See how an always-on operator works.

Hand off your first task tonight.

Tell us your email and what you'd hand off first. We'll send your access details and help you set up your operator.