Context Window Calculator
A 100,000-token text fills 10% of a 1,000,000-token context window and leaves 900,000 tokens.
How to use
- Paste your text or load a text file. Or leave the box empty and type a token count.
- Optionally set aside tokens for the answer in the reserve field.
- Read the table: percent used, tokens left and the rough words and pages left for each model. Rows turn red when the text does not fit.
How it works
Percent used = tokens ÷ context window × 100. Tokens left = context window − reserved tokens − tokens used. A text fits when tokens left is zero or more.
Tokens come from your text. GPT-4o and GPT-4.1 use the exact o200k_base count (js-tiktoken). The GPT-6 models have no tokenizer name in OpenAI's docs, so their rows use o200k_base as a reference (marked †). Claude and Gemini rows are estimates (marked ‡) from the ratios in Anthropic's and Google's documentation: Claude 2.5 characters or 0.555 words per token, Gemini 4 characters or 60 to 80 words per 100 tokens, using the larger value. If you type a token count instead, the same number is used for every model.
Words left = tokens left × words per token. With pasted text, words per token is measured from your text for each model; with a typed token count it is the documented ratio (Claude 0.555, Gemini 0.7, the middle of 60 to 80 per 100) or 0.75 for OpenAI, a common rule of thumb. Pages left = words left ÷ words per page, 500 by default, which you can change.
Models, limits and sources
Context window and maximum output are the figures printed on each provider's documentation page, checked on 2026-10-15.
Windows can change or differ by plan and platform. Check the source page before relying on a number.
| Model | Context window (tokens) | Max output (tokens) | Source (checked) |
|---|---|---|---|
| Claude Fable 5.1 (claude-fable-5-1) | 1,000,000 | 128,000 | https://platform.claude.com/docs/en/about-claude/models/overview (2026-10-15) |
| Claude Opus 5.5 (claude-opus-5-5) | 1,000,000 | 128,000 | https://platform.claude.com/docs/en/about-claude/models/overview (2026-10-15) |
| Claude Sonnet 5.5 (claude-sonnet-5-5) | 1,000,000 | 128,000 | https://platform.claude.com/docs/en/about-claude/models/overview (2026-10-15) |
| Claude Haiku 5.5 (claude-haiku-5-5) | 1,000,000 | 128,000 | https://platform.claude.com/docs/en/about-claude/models/overview (2026-10-15) |
| GPT-6 Astra (gpt-6-astra) | 1,050,000 | 128,000 | https://developers.openai.com/api/docs/models/gpt-6-astra (2026-10-15) |
| GPT-6.1 Sol (gpt-6.1-sol) | 1,050,000 | 128,000 | https://developers.openai.com/api/docs/models/gpt-6.1-sol (2026-10-15) |
| GPT-6 Luna (gpt-6-luna) | 1,050,000 | 128,000 | https://developers.openai.com/api/docs/models/gpt-6-luna (2026-10-15) |
| GPT-4.1 (gpt-4.1) | 1,047,576 | 32,768 | https://developers.openai.com/api/docs/models/gpt-4.1 (2026-10-15) |
| GPT-4o (gpt-4o) | 128,000 | 16,384 | https://developers.openai.com/api/docs/models/gpt-4o (2026-10-15) |
| Gemini 3.8 Flash (gemini-3.8-flash) | 1,048,576 | 65,536 | https://ai.google.dev/gemini-api/docs/models/gemini-3.8-flash (2026-10-15) |
| Gemini 3.1 Pro (preview) (gemini-3.1-pro-preview) | 1,048,576 | 65,536 | https://ai.google.dev/gemini-api/docs/models/gemini-3.1-pro-preview (2026-10-15) |
| Gemini 2.5 Pro (gemini-2.5-pro) | 1,048,576 | 65,536 | https://ai.google.dev/gemini-api/docs/models/gemini-2.5-pro (2026-10-15) |
| Gemini 2.5 Flash (gemini-2.5-flash) | 1,048,576 | 65,536 | https://ai.google.dev/gemini-api/docs/models/gemini-2.5-flash (2026-10-15) |
FAQ
How many words fit in a 1M-token context window?
Anthropic says 1M tokens is roughly 555k words on its current tokenizer. Google says 100 tokens are about 60 to 80 English words, so 1M tokens is about 600k to 800k words. It varies with the language and the content.
Does the context window include the answer?
The provider docs define how input, output and reasoning tokens share the window. Use the reserve field to set aside the tokens you expect in the answer. The table in the guide lists each model's maximum output.
How accurate is the percentage?
The OpenAI GPT-4o and GPT-4.1 numbers are exact for the text alone. GPT-6 is a reference, and Claude and Gemini are estimates, so keep a safety margin near the limit and confirm with the provider's token counting API.
Why is my favorite model missing?
Only models whose context window was confirmed on an official documentation page are listed, with the source and check date in the guide. Limits change and some plans or platforms offer smaller windows, so verify on the linked page before you rely on a number.
Is my text sent anywhere?
No. Everything runs in your browser. Files you load are read locally and never uploaded.