◈AI Codex
Foundation Models & LLMsCore Definition

The whiteboard every AI conversation shares

In brief

Context window is the single number that shapes everything about how Claude thinks with you — and most people are using only a fraction of it.

5 min read·Context Window

Contents

♡Sign in to save

Think of a context window like a whiteboard.

Every time you start a conversation with Claude, a blank whiteboard appears. Everything goes on that whiteboard — your instructions, the documents you share, the questions you ask, Claude's answers, the follow-ups. When the whiteboard fills up, the oldest things get erased to make room for new ones.

The context window is the size of that whiteboard. Measured in tokens (roughly ¾ of a word each), it determines how much Claude can hold in mind at once during a conversation.

Why this matters more than most people realize

Most AI assistants have historically had small whiteboards — 4,000 to 8,000 tokens. Enough for a back-and-forth conversation, not much more.

Claude's context window changes the game. Claude Fable 5, Claude Opus 5 and Claude Sonnet 5 each support 1,000,000 tokens — roughly 555,000 words on the current tokenizer. That's several novels. Or a company's full legal documentation. Or a year of customer support tickets. All available to Claude at once, without losing the thread. Claude Haiku 4.5, the fast model, carries 200,000 tokens.

This isn't just a bigger number. It's a different kind of tool.

With a small context window, you have to be strategic about what you tell the model. You summarize, compress, select. With a million tokens, you can often just give Claude everything and let it find what matters.

What fits in 1,000,000 tokens

To make this concrete:

  • A full product specification document: fits, using a fraction of a percent
  • An entire codebase for a mid-sized application: fits
  • All your customer interviews from a discovery sprint: fits
  • Several years of email threads with a key partner: fits
  • The complete works of Shakespeare: fits about six times over

The practical ceiling stopped being "will it fit" and became "does including it help." Those are different questions, and the second one is now the one that matters.

How Claude handles the whiteboard

When Claude reads a long document inside its context window, it doesn't skim. It processes the full text. This means Claude can answer questions about page 147 with the same accuracy as page 3 — as long as both are in the window.

The whiteboard is still per-conversation: what is on it is what Claude is actively reading. Claude does now carry memory between conversations, but that works differently — memory is a set of retained entries that get brought to the whiteboard, not a whiteboard that never gets wiped. For information too large or too fast-changing to sit in either, that's where RAG comes in.

The practical move

Most people use Claude like a search engine — quick questions, short exchanges. The context window makes a different workflow possible: load everything once, ask many things.

If you're analyzing a contract, paste the whole contract. If you're onboarding Claude into a project, dump all the relevant docs up front. If you're debugging a system, share the full logs. Claude will handle it.

The whiteboard is big. Use it.

Further reading

Official training on this: AI Capabilities and Limitations (13 lessons · 3.5 hr) · Parametric memory and context on Claude Academy, free.

Weekly brief

For people actually using Claude at work.

Each week: one thing Claude can do in your work that most people haven't figured out yet — plus the failure modes to avoid. No tutorials. No hype.

No spam. Unsubscribe anytime.

What to read next

All articles →
Context Window·Role-Specific·4 min

How to think about Claude's context window (and when it actually matters)

A million tokens sounds like more room than anyone could use. In practice, how you use that space — not whether you run out of it — is what changes the quality of your outputs.

Context Window·How It Works·5 min

The context window in practice: what it means for how you work

The context window shapes what Claude can and can't do in any given conversation. Here is how to work with it.

How It Works·14 min

Claude Fable 5.1: cache reads at a quarter of the price, and three changes that break existing code

On September 1, 2026 Anthropic shipped Claude Fable 5.1 at the same $10/$50 per million tokens as Fable 5, with cache reads cut from $1.00 to $0.25 per million. That single line makes long agent sessions roughly 25% cheaper, and up to 45% cheaper for heavily agentic work. Three things break on migration: forced tool use, cross-model thinking blocks, and any code that edits earlier turns. Here is what changed and what to check before you switch the model ID.