Your session has been flagged for unusual activity.
Please confirm you're human to continue.
This context window visualizer shows how much of each AI model's context window your text fills, for GPT-6, Claude, Gemini and Grok. Paste a prompt or document, or enter a token count, and plan long-context work before you hit the limit.
Est. tokens
0
Characters
0
≈ Words
0
≈ Pages
0
Tokens are estimated at ~4 characters per token. Words ≈ tokens × 0.75; pages at 250 words each.
API context windows from each provider's model docs, October 2026. Chat apps often use a smaller window; see app vs API.
A context window is the most text an AI model can take into account at once. An LLM context window is measured in tokens, not words or characters: about 4 characters, or three-quarters of a word, in English. The glossary entry for context window goes into more depth.
Everything in a request shares the same window:
Instructions
The system prompt, custom instructions and any tool definitions.
Conversation history
Every earlier message is sent again with each new turn.
Files and documents
Pasted text, uploaded PDFs, code and search results.
The answer
The model's response also uses tokens from the same window.
When the window is full: on the API, a request that is too long is rejected with an error. In chat apps, a conversation that outgrows the window can no longer be taken into account in full. Starting a new chat with a short summary of the old one usually fixes it.
The maximum context window and longest single answer each model supports on its API. Almost every current flagship now offers about 1M tokens; Grok 4.7 is the exception at 500K.
| Model | Context window | Max output | ≈ Words |
|---|---|---|---|
| GPT-6 AstraOpenAI | 1,050,000 | 128,000 | ~788K |
| GPT-6.1 SolOpenAI | 1,050,000 | 128,000 | ~788K |
| GPT-6 LunaOpenAI | 1,050,000 | 128,000 | ~788K |
| Claude Fable 5.1Anthropic | 1,000,000 | 128,000 | ~750K |
| Claude Opus 5.5Anthropic | 1,000,000 | 128,000 | ~750K |
| Claude Sonnet 5.5Anthropic | 1,000,000 | 128,000 | ~750K |
| Claude Haiku 5.5Anthropic | 1,000,000 | 128,000 | ~750K |
| Gemini 4 ArgonGoogle | 1,048,576 | 65,536 | ~786K |
| Gemini 3.8 FlashGoogle | 1,048,576 | 65,536 | ~786K |
| Grok 4.7xAI | 500,000 | — | ~375K |
| Grok 4.3xAI | 1,000,000 | — | ~750K |
Tokens. Sources: OpenAI API models docs, Anthropic models overview, Google Gemini API model pages and xAI models docs, checked October 2026. Gemini 4 Argon writes 65,536 tokens per answer by default and up to its full window with long decode. "—" means the provider doesn't list a separate output cap.
A 1M-token window means a model can read a whole book or codebase. It doesn't mean that's the best way to use it.
Quality can drop in very long prompts
Models tend to use information at the start and end of a long prompt more reliably than details buried in the middle, a pattern documented in the "Lost in the Middle" study. Put key instructions at the top or bottom, and send only the material the task needs.
You pay for every token, on every call
On the API, the whole context is billed as input each time you send it. A 500,000-token document costs $2.00 per request on Claude Opus 5.5 ($4 per 1M input tokens) before the model writes a word, and follow-up questions resend it. Estimate it with the AI cost calculator.
Long prompts are slower
The model has to process every input token before it answers, so a near-full window means a noticeably longer wait for the first word.
Rule of thumb: keep headroom for the answer and the next few turns. If the visualizer shows a bar turning amber (70%+) or red (90%+), trim the input, split the task, or start a fresh chat with a summary.
The numbers above are API maximums. In the consumer apps, the context window depends on your plan, and for ChatGPT it is much smaller than the API's 1.05M tokens.
| App and plan | Context in the app |
|---|---|
| ChatGPT Free | About 27K |
| ChatGPT Plus ($20/month) | 54K Instant · 256K reasoning |
| ChatGPT Pro | 128K Instant · 400K reasoning |
| Claude Pro and Max | 1M |
| Google AI Pro ($19.99/month) | 1M |
Tokens. From chatgpt.com/pricing, OpenAI and Anthropic help centers and Google's Gemini plan pages, checked October 21, 2026. Plans and limits change often; see our ChatGPT and Claude limits guide for message caps by plan.
What this means in practice: a 150,000-token contract fits comfortably in Claude, Gemini or any API model, but not in a ChatGPT Plus Instant chat. In apps, uploaded files can also be handled differently from pasted text, so the visualizer is most accurate for text you paste directly or send through an API.
A context window is the maximum amount of text an AI model can take into account at once, measured in tokens. It covers everything in the request: the system prompt, the conversation so far, any files or documents you add, and the response the model writes.
On the APIs in October 2026 it is close to a tie: GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna list 1,050,000 tokens, Gemini 4 Argon and Gemini 3.8 Flash 1,048,576, the Claude 5.5 models and Claude Fable 5.1 1,000,000, and Grok 4.3 1,000,000. Grok 4.7 lists 500,000.
Roughly 750,000 words of English, or about 3,000 pages at 250 words per page, using the common estimate of 0.75 words per token. Code, tables and non-English text use more tokens per word, so less of it fits.
On the API, a request that is longer than the model's context window is rejected with an error, so you have to shorten it. Chat apps handle it differently: when a long conversation outgrows the window, the model can no longer take all of it into account, and starting a new chat with a summary of the old one usually helps.
No. The OpenAI API lists 1,050,000 tokens for GPT-6 Astra, but in the ChatGPT app the window depends on your plan: about 27K tokens on Free, 54K for Instant and 256K for reasoning models on Plus, and 128K and 400K on Pro. Paid Claude plans and Google AI Pro offer a 1M-token window in the app.
Yes. The window is shared by your input and the model's output. The max output limit is a separate, smaller cap on how long a single answer can be, for example 128,000 tokens for GPT-6 and Claude 5.5 models and 65,536 by default for Gemini 4 Argon.
It estimates tokens at about 4 characters per token, which is close for typical English text. Each provider uses its own tokenizer, so real counts can differ by 10 percent or more, especially for code and other languages. For an exact GPT count, use the token counter, which runs OpenAI's tokenizer in your browser. Your text is never sent anywhere.
No. Every token you send is billed and slows down the answer, and models tend to use information at the start and end of a very long prompt more reliably than details buried in the middle. Send what the task needs, keep headroom, and put key instructions where the model will see them clearly.
MultipleChat puts ChatGPT, Claude, Gemini and Grok in one workspace, so a long document can go to the model that handles it best, and you can compare the answers side by side.
Try MultipleChat