Free tools/Context Window Comparison

Compare model context windows

Tokens · approximate words · approximate pages

Compare the stored context limits for the listed models, with rough word and page equivalents.

GPT-6 Luna OpenAI1,050k tokens (input + output) · ~1,575 pages
GPT-6.1 Sol OpenAI1,050k tokens (input + output) · ~1,575 pages
Gemini 3.8 Flash Google1,048.576k tokens (input) · ~1,573 pages
Claude Haiku 5.5 Anthropic1,000k tokens (input + output) · ~1,500 pages
Claude Sonnet 5.5 Anthropic1,000k tokens (input + output) · ~1,500 pages
Claude Opus 5.5 Anthropic1,000k tokens (input + output) · ~1,500 pages

Largest here: GPT-6 Luna at 1,050k tokens — roughly 787,500 words, or about 1,575 pages of text as a capacity illustration, before reserving output space.

Context is the provider-published capacity; it may include output as well as input. Reserve room for output and instructions. Word/page estimates assume 0.75 words per token and 500 words per page. Sources checked 2026-11-12.

How to use it

  1. Estimate the full request

    Count the prompt, instructions, previous messages, and documents needed for the task.

  2. Compare the available limits

    Use the bars to compare listed models. Page estimates are approximate and vary with language and formatting.

  3. Check the provider’s details

    Confirm input and output limits for the exact model. Test long documents for answer quality as well as size.

More questions

Are page counts exact?

No. They use rough token-to-word and word-to-page conversions. Code, tables, and other languages can differ substantially.

Should I always use the largest window?

Choose the context your task needs. Retrieving relevant sections can be more useful than sending every available document.

Method and example

The chart shows the provider-published token capacity. Some providers define a combined input/output window; others publish a separate input limit.

Illustrative example: A 100,000-token capacity is roughly 75,000 English words using the display assumption of 0.75 words per token. This is not a measured tokenizer result.

Allow space for output and system instructions. A document fitting into context does not guarantee that every detail will be retrieved correctly.

More free AI tools.

Discuss your result with us

Bring the result and one workflow to a free 30-minute call. We will review the assumptions and discuss a practical next step.