Context windows · July 2026
A context window is the total amount of text (measured intokens) a model can read and write in a single request. Bigger windows mean you can feed in longer documents, entire codebases, or whole books — and still get a response. Here is how the major models compare in 2026.
Values calculated from each model's context window using the standard ratio of 1 token ≈ 0.75 English words, 500 words per page, and ~90,000 words (120K tokens) per novel.
| Model | Context | Words | Pages | Novels | Equivalent |
|---|---|---|---|---|---|
| GPT-5.6 Sol | 1.05M | 787,500 | 1,574 | 8.8 | ~8.75 novels — the Bible + Harry Potter |
| GPT-5.6 Terra | 1.05M | 787,500 | 1,574 | 8.8 | ~8.75 novels — the Bible + Harry Potter |
| GPT-5.6 Luna | 1.05M | 787,500 | 1,574 | 8.8 | ~8.75 novels — the Bible + Harry Potter |
| Claude Fable 5 | 1M | 750,000 | 1,499 | 8.3 | ~8 novels — nearly the full Bible |
| Claude Opus 4.8 | 1M | 750,000 | 1,499 | 8.3 | ~8 novels — nearly the full Bible |
| Claude Sonnet 5 | 1M | 750,000 | 1,499 | 8.3 | ~8 novels — nearly the full Bible |
| Gemini 3.1 Pro | 1M | 750,000 | 1,499 | 8.3 | ~8 novels — nearly the full Bible |
| Gemini 3.5 Flash | 1M | 750,000 | 1,499 | 8.3 | ~8 novels — nearly the full Bible |
| Gemini 3 Flash | 1M | 750,000 | 1,499 | 8.3 | ~8 novels — nearly the full Bible |
| GPT-5.5 Pro | 400K | 300,000 | 600 | 3.3 | ~3 novels — one War and Peace |
| GPT-5.5 | 400K | 300,000 | 600 | 3.3 | ~3 novels — one War and Peace |
| Claude Haiku 4.5 | 200K | 150,000 | 300 | 1.7 | ~1.7 novels |
1.05M tokens
GPT-5.6 Sol, GPT-5.6 Terra, GPT-5.6 Luna
1M tokens
Claude Fable 5, Claude Opus 4.8, Claude Sonnet 5, Gemini 3.1 Pro, Gemini 3.5 Flash, Gemini 3 Flash
400K tokens
GPT-5.5 Pro, GPT-5.5
200K tokens
Claude Haiku 4.5
The context window is the maximum number of tokens a model can process in a single request, including both your input (the prompt, documents, images you send) and the model's output (the response it generates). Think of it as the model's working memory — everything it can "see" at once.
As of July 2026, GPT-5.6 Sol, GPT-5.6 Terra, GPT-5.6 Luna have the largest context windows at 1.05M tokens (1,050,000 tokens). That fits roughly 787,500 words — equivalent to about 8.8 average novels.
Not entirely for input, because the model's output tokens count against the context window too. If a model has a 1M-token context window and you need a 2,000-token response, your input is effectively capped at 998,000 tokens. Most providers also set a separate maximum output length (typically 8K–32K tokens) that is smaller than the full window.