The Claude Haiku 4.5 context window is 200,000 tokens, enough for about 133,333 English words: 267 pages, or 1.7 80,000-word books.
| Item | Value |
|---|---|
| Words | 133,333 |
| Pages | 267 |
| Books (80K words) | 1.7 |
| Lines of code | 15,500 |
| Hours of transcribed speech | 16 h |
| Cost to fill the context (input) | $0.20 |
In practice, leave room for the answer and the history: the context is shared between input and output.
Yes. Input and output share one window, and output has its own cap (128K tokens on the larger models).
Only when the model needs to see everything at once. For finding passages across many documents, RAG is usually cheaper and more accurate.
Estimate based on public prices as of October 2026. Token counts vary with the text and the model; check the official pricing page before you decide.