What is a Context Window in AI? | AI Glossary

Intermediate

TLDR

A context window is the total amount of text an AI model can hold in its working memory at once, including everything said in the current conversation and any documents you have shared.

Think of an AI's context window as its short-term memory. Everything within the window is something the model can actively reference when generating its next response. Once the conversation exceeds the context window, the oldest content starts to drop out, and the model can no longer "see" or refer to it.

Context window size is measured in tokens, where one token is roughly three to four characters of English text. A model with a 128,000-token context window can hold roughly 100,000 words at once: the equivalent of a short novel. A model with a 200,000-token window (like Claude) can handle even more. Earlier models from 2022 and 2023 had windows as small as 4,000 tokens, which severely limited what you could ask them to do in a single session.

Context window size determines what tasks an AI can handle in a single session. Summarizing a 50-page report, reviewing a full codebase, or maintaining consistency across a long document all require a large context window. For most everyday tasks, modern context windows are large enough that hitting the limit is rare. For power users working with long documents or complex projects, it remains a practical constraint.

In practice

Document analysis

Uploading a 30-page research report and asking the AI to summarize it, extract key insights, or answer specific questions requires the entire document to fit within the context window.

Long coding sessions

When debugging a large codebase, you may need to paste multiple files into the context. A model with a small context window will lose track of earlier files as the conversation grows.

Multi-chapter editing

Asking an AI to edit a 20,000-word manuscript for consistency in voice and character detail requires a large enough context window to hold the entire document at once.

Related terms

Tokens Prompt Engineering System Prompt

Frequently asked questions

What happens when you exceed the context window?+

When a conversation exceeds the context window, the model can no longer see the oldest messages. It will continue responding but may lose track of instructions given early in the session, contradict earlier statements, or seem to "forget" constraints you set at the start. For long sessions, this is why re-stating key instructions periodically helps.

Which AI model has the largest context window?+

As of 2026, Claude from Anthropic offers one of the largest context windows at 200,000 tokens (roughly 150,000 words). Gemini 1.5 Pro supports up to 1 million tokens in some configurations. Context window sizes continue to increase rapidly across all major models.

Does a larger context window mean better quality?+

Not necessarily. A larger context window means the model can handle more text at once, but it does not guarantee better reasoning or writing quality. Some research suggests that models can lose focus on information from the middle of very long contexts. Context window size and output quality are separate dimensions of model performance.

How many pages fit in a 100,000-token context window?+

Roughly 75,000 words, which is approximately 300 standard pages of text. This is enough to hold a full-length business book, a long legal contract, or a large software project with multiple files. The exact count varies by language and formatting.

Bottom line

A context window is the total amount of text an AI model can hold in its working memory at once, including everything said in the current conversation and any documents you have shared.