Free tool
Context Window Calculator
Paste your prompt and instantly see how much of each model's context window it uses — GPT-4o, Claude, Gemini and more — including room reserved for the response.
GPT-4o
3.1% used
GPT-4o mini
3.1% used
Claude (Sonnet / Opus / Haiku)
2.0% used
Gemini Pro
0.4% used
Gemini Flash
0.4% used
Llama 3.1
3.1% used
Mistral Large
3.1% used
DeepSeek
3.1% used
Context windows shown are common defaults and change between model versions — treat this as a planning estimate.
What it is for
Every model has a ceiling on how much it can consider at once, and everything counts against it — your prompt, the documents you paste, the conversation so far, and the reply it is about to write.
Exceeding it does not usually produce an error. It produces silence about the parts that were dropped, which is considerably worse.
How it works
Paste your text and see what proportion of each major model's window it occupies, with room reserved for the response. Models are compared side by side so the question "will this fit anywhere" gets a direct answer.
Estimates use the same approximation as the token counter — good to roughly 10–15% for English prose, less reliable for code and non-English text.
When to use it — and when not to
Use it before pasting something long, when choosing a model for a document-heavy task, and when deciding whether you need retrieval or can simply hand the model everything.
That last decision is worth taking seriously. Retrieval pipelines exist largely because context used to be expensive; for a small corpus, putting the whole thing in the prompt is often simpler and more accurate than building infrastructure to feed the model less.
Things worth knowing
A large window is not the same as good use of a large window. Accuracy on material buried in the middle of a very long prompt degrades measurably across every model — the effect has a name, "lost in the middle". Put the important material at the start or the end, and verify rather than assume when a fact came from the middle of a hundred pages.
Reserve at least a quarter of the window for the response. A prompt that fits perfectly leaves nowhere for the answer to go.
Frequently asked questions
What is a context window?
The context window is the maximum amount of text — measured in tokens — a model can consider at once, including your prompt, any documents, the conversation history and its own response. Exceed it and the earliest content is dropped.
Why reserve tokens for the response?
The context window covers input and output together. If you fill it entirely with your prompt, the model has no room left to answer. Reserving 2,000-4,000 tokens is a sensible default.