Context window sizes by model

The context window is how much text a model can read and write in one request. This table lists the window and maximum output for the current Claude, GPT, Gemini, DeepSeek and Grok models from each provider's own model page, and is available as a CSV. Measure your own text in the token counter; the rest of the topic is in tokens and context windows.

By the Chelure Enterprises Editorial Team. Last reviewed 13 November 2026. How we check our sources.

Context window and maximum output

Download as CSV (with the source URL for each row). Checked November 13, 2026. Mistral's models overview page did not state context windows for Large 4, Small 4 or Ministral, so they are not listed; add them when the provider publishes the figure.

How many words fit?

Anthropic states that 200K tokens is roughly 150K words, and 1M tokens roughly 555K words on its current tokenizer (introduced with Claude Opus 4.7); models before it fit about 750K words in 1M tokens. For other text use the tokens to words converter or paste it into the counter, because code, tables and non-English text use more tokens per word.

What a bigger window does and does not do

A larger window lets you include more material in a single request, but you pay for every input token on every request, so a 500K-token prompt sent repeatedly is expensive unless it is cached (see prompt caching). Some prices also step up for long prompts; the model pages note where. Use the cost calculator to price a long prompt before you send it.

Sources

• Anthropic: models overview (context window, max output, words per token)
• OpenAI: models (GPT-6 family)
• Google: Gemini models
• DeepSeek: models, limits and pricing
• xAI: models

Primary sources opened on November 13, 2026. Prices come from our verified price table (see the methodology).