Explore a 1M-token context window
Add books and code to an estimated token budget. Leave room for instructions, tool results and output.
📖 Historical article: Sonnet 4.6, two experiments, one got personal (February 2026) →Seven novels. Approximate English word count, converted using the assumption below.
Estimated: 1,445,560 tokens
The trilogy only, excluding The Hobbit. Approximate English word count.
Estimated: 630,968 tokens
Approximate English word count. Translation and edition change the total.
Estimated: 773,334 tokens
Old and New Testament. Approximate English word count; the estimated tokens exceed 1M.
Estimated: 1,044,183 tokens
One short message. An illustrative budget; length and language vary.
Estimated: 15 tokens
One medium email. An illustrative budget, without attachments.
Estimated: 200 tokens
A meeting transcript. An illustrative text budget, excluding audio.
Estimated: 8,000 tokens
A long podcast transcript. An illustrative text budget, excluding audio.
Estimated: 20,000 tokens
A synthetic large repository budget, not a measured Linux release.
Estimated: 25,000,000 tokens
React v19.2.0, packages/react: 52 JavaScript files, excluding __tests__. Measured source characters, estimated tokens.
Estimated: 33,953 tokens
CPython v3.13.0, Lib: 695 Python files, excluding Lib/test. Measured source characters, estimated tokens.
Estimated: 2,974,941 tokens
A synthetic small project budget. Excludes build output, dependencies and images.
Estimated: 200,000 tokens
One journal entry. An illustrative budget for a few paragraphs.
Estimated: 500 tokens
One article. An illustrative text budget, not a sampled Wikipedia average.
Estimated: 1,500 tokens
One collection of news articles. An illustrative budget, not all global coverage.
Estimated: 50,000 tokens
One short story. About 450 English words under the same estimate.
Estimated: 600 tokens
What the budget tells you: A larger window can hold more source text. It still needs room for your instructions, tool results and output. Test whether the model uses the details your task depends on.
Source releases: React v19.2.0 · CPython v3.13.0
Estimates assume 0.75 English words per token, or 4 characters per token for source text. Actual tokenization varies by model, language and content.
Book word counts are approximate inputs, not measured token counts. React core and Python stdlib use source characters counted on 2026-09-13. Other entries are illustrative budgets, not measured corpus sizes.