Free tool · Context windows
Context window checker
Paste any text and see which of the 19 models in llmwise can read it in one go: its token count, the share of each model's context window it fills, and whether a reply still fits. Free, in your browser.
… tokens by o200k_base (0 by llmwise's estimate of 4 characters a token). Paste something to see which models it fits.
A chat in llmwise holds up to 200,000 tokens, whatever the model's window.
| Model | Context window | This text uses | Fits, with a reply |
|---|---|---|---|
| GPT-6 AstraOpenAI | 1,050,000 | 0.0% | Yes |
| GPT-6.1 SolOpenAI | 1,050,000 | 0.0% | Yes |
| GPT-6 SolOpenAI | 1,050,000 | 0.0% | Yes |
| GPT-6 LunaOpenAI | 1,050,000 | 0.0% | Yes |
| Gemini 3.1 ProGoogle | 1,048,576 | 0.0% | Yes |
| Gemini 3.8 FlashGoogle | 1,048,576 | 0.0% | Yes |
| DeepSeek V4.1 FlashDeepSeek | 1,048,576 | 0.0% | Yes |
| DeepSeek V4 ProDeepSeek | 1,048,576 | 0.0% | Yes |
| Kimi K3Moonshot | 1,048,576 | 0.0% | Yes |
| GLM 5.3Z.ai | 1,048,576 | 0.0% | Yes |
| GLM 5.3 FlashZ.ai | 1,048,576 | 0.0% | Yes |
| Mistral Large 4Mistral | 1,048,576 | 0.0% | Yes |
| Claude Fable 5.1Anthropic | 1,000,000 | 0.0% | Yes |
| Claude Opus 5.5Anthropic | 1,000,000 | 0.0% | Yes |
| Claude Sonnet 5.5Anthropic | 1,000,000 | 0.0% | Yes |
| Claude Sonnet 5Anthropic | 1,000,000 | 0.0% | Yes |
| Claude Haiku 5.5Anthropic | 1,000,000 | 0.0% | Yes |
| Grok 4.7xAI | 500,000 | 0.0% | Yes |
| Claude Haiku 4.5Anthropic | 200,000 | 0.0% | Yes |
Counted in your browser with OpenAI's o200k_base tokenizer: exact for that tokenizer, an estimate for every model here, since each counts with its own. Your text isn't sent anywhere.
Context windows checked . Updated .
Every model's context window, in words and pages
Each model's context window, and roughly how much English text that is at the 1.15 tokens a word we measured (a page is about 500 words).
| Model | Tokens | About this many words | About this many pages |
|---|---|---|---|
| GPT-6 Astra | 1,050,000 | 909,000 | 1,818 |
| GPT-6.1 Sol | 1,050,000 | 909,000 | 1,818 |
| GPT-6 Sol | 1,050,000 | 909,000 | 1,818 |
| GPT-6 Luna | 1,050,000 | 909,000 | 1,818 |
| Gemini 3.1 Pro | 1,048,576 | 908,000 | 1,816 |
| Gemini 3.8 Flash | 1,048,576 | 908,000 | 1,816 |
| DeepSeek V4.1 Flash | 1,048,576 | 908,000 | 1,816 |
| DeepSeek V4 Pro | 1,048,576 | 908,000 | 1,816 |
| Kimi K3 | 1,048,576 | 908,000 | 1,816 |
| GLM 5.3 | 1,048,576 | 908,000 | 1,816 |
| GLM 5.3 Flash | 1,048,576 | 908,000 | 1,816 |
| Mistral Large 4 | 1,048,576 | 908,000 | 1,816 |
| Claude Fable 5.1 | 1,000,000 | 866,000 | 1,732 |
| Claude Opus 5.5 | 1,000,000 | 866,000 | 1,732 |
| Claude Sonnet 5.5 | 1,000,000 | 866,000 | 1,732 |
| Claude Sonnet 5 | 1,000,000 | 866,000 | 1,732 |
| Claude Haiku 5.5 | 1,000,000 | 866,000 | 1,732 |
| Grok 4.7 | 500,000 | 433,000 | 866 |
| Claude Haiku 4.5 | 200,000 | 173,000 | 346 |
The windows run from 200,000 tokens (Claude Haiku 4.5) to 1,050,000 (GPT-6 Astra). Words and pages are rough: they assume plain English prose.
Long texts in llmwise
A chat holds up to 200,000 tokens, whatever the model's window: past that, start a new chat or attach the document and ask about parts of it.
Past 64,000 tokens a reply counts as 2 messages on your plan, and past 128,000 as 4; the message box says so before you send.
The checker keeps 16,000 tokens for the reply, the longest reply a paid plan allows.
Attached PDFs are read as text or as page images depending on the model, which changes how many tokens they take; the message box shows the estimate before you send.
For what a context window is and what fills it, read what is a context window; to price a text on every model, use the token counter.
Questions
Which model has the largest context window?
GPT-6 Astra's, at 1,050,000 tokens: about 909,000 words of English by our measure. In llmwise a chat holds up to 200,000 tokens whatever the model's window, so a longer document is best attached as a file.
How many words fit in a context window?
About 0.87 English words a token, going by our sample (71 words, 82 tokens with OpenAI's o200k_base tokenizer). Code and other languages take more tokens a word, so the same window holds fewer words of them.
What counts toward the context window?
Everything the model reads for the reply: the conversation so far, any files or pasted text, the instructions the app adds, and the reply itself as it's written. That's why the checker keeps room for a reply.
What happens when text doesn't fit?
The provider refuses the request, or the app has to drop the oldest messages. In llmwise a chat takes no new message past 200,000 tokens, and a reply counts as 2 messages past 64,000 and 4 past 128,000, which the message box shows before you send.
Is the count exact for Claude or Gemini?
Close, not exact: the count uses OpenAI's o200k_base tokenizer, and Claude, Gemini, Grok and the open-weight models count with their own. Leave some margin when a text is near a model's limit.
Claude, GPT, Gemini, DeepSeek, Grok, Kimi, GLM, and Mistral, in one chat.
See what a message costs before you send it. Free is 5 messages to try; sign in with an email link, no password or card.