What is a context window? (And why it matters for your business)

You have probably seen the phrase "context window" mentioned in AI news and quietly moved past it. It sounds like a technical detail. It is actually one of the most practical things to understand about how AI works, and recent changes to context window sizes have made AI tools meaningfully more useful for business.

Here is the plain English version.

Working memory

The best analogy for a context window is working memory. When you are in a meeting, your brain holds onto what was said five minutes ago, connects it to what someone just said now, and uses both to inform what you are about to say. If the meeting goes on long enough, your working memory starts to lose the earlier parts. You might forget the specific point that was made at the start of the session.

AI has the same limitation, scaled differently. Every AI conversation exists inside a context window. That window holds everything: your prompts, the AI's responses, any documents you have uploaded, everything. The AI can only work with what is inside the window. When something falls outside it, the AI cannot see or reference it.

A small context window means short-term forgetting. A large one means the AI can hold much more before anything slips away.

What tokens are

Context window sizes are measured in tokens, not words. A token is roughly three to four characters, or about three quarters of a word on average. One million tokens is approximately 750,000 words.

To give that some scale: the average business book runs to around 70,000 to 80,000 words. A million token context window could hold roughly ten of those books at once.

For most business tasks, you will use a fraction of that. But the ceiling matters because it removes a practical constraint you may not have even noticed was there.

Why it matters in practice

Longer documents without chopping them up. Previously, if you wanted to feed a long report or contract to an AI, you might have had to break it into sections, losing the coherence of the whole. A large context window means you can upload the entire document and ask questions about it as a whole.

Longer conversations without the AI forgetting. If you have ever had a long AI chat session where the assistant seemed to forget something from earlier in the conversation, that was the context window at work. A larger window means more of the conversation stays available to the AI throughout.

Bigger projects in a single session. For developers or anyone working with large files, a 1 million token context window means an entire codebase or full dataset can be read in one pass rather than piecemeal.

More coherent output across complex tasks. When the AI can hold the whole picture, its answers are more consistent. It does not contradict itself halfway through a long document because it has lost track of what it said at the start.

What just changed

Claude, from Anthropic, recently extended its context window to 1 million tokens and made it available to all users, including on the free tier. This is a meaningful step. Most AI tools have had context windows of 128,000 tokens or fewer, which is still considerable but noticeably limiting for large documents or long projects.

Other AI tools are moving in the same direction. Context window sizes have been growing steadily and will continue to do so.

The practical takeaway

You do not need to manage context windows yourself. The main thing to know is that bigger is better, that Claude currently leads on this, and that the 1 million token window removes most of the practical limits you might have hit before.

If you have a large document you want to analyse, a long thread of emails you want summarised, or a complex project you want to work through in detail, try Claude. Upload the whole thing. See what comes back.

The tool is ready for it.