The detail
Most adult novels run 70,000–100,000 words, so call the average about 90,000 words, or roughly 120,000 tokens. Non-fiction and longer fantasy titles can reach 150,000 words (about 200,000 tokens), and epics go further still.
This is exactly where large context windows changed what is possible: a 200,000-token model can hold most full novels in a single prompt, and million-token models can take several books at once — enabling whole-book summarization, search, and analysis without chunking.
Counting a full book precisely is worthwhile before a large run, because cover matter, footnotes, and indices add up, and at 120,000 input tokens per request the cost is no longer trivial. Use the pricing table to compare models before processing long texts.