Context window comparison, every major model.
How much each model can actually hold: the context window in tokens, what that means in plain words, and the API input price next to it. Numbers come from the providers' own documentation, last re-verified 2026-07-07. Counting a specific prompt instead? Use the token counter.
| model | context window | ≈ words that fit | api input / 1m |
|---|---|---|---|
| Llama 4 Scout Meta 10M advertised by Meta; most hosts serve 128K to 1M | 10M | ~7.5M | — |
| GPT-5.5 OpenAI very long prompts bill at a higher tier | 1.1M | ~788K | $5 |
| Gemini 2.5 Pro Google $1.25 up to 200K-token prompts, $2.50 beyond | 1.0M | ~786K | $1.25 |
| Gemini 2.5 Flash Google | 1.0M | ~786K | $0.3 |
| GPT-4.1 OpenAI | 1.0M | ~786K | $2 |
| Claude Opus 4.8 Anthropic no long-context surcharge | 1M | ~750K | $5 |
| Claude Sonnet 5 Anthropic | 1M | ~750K | $3 |
| Gemini 3.1 Pro Google $2 up to 200K-token prompts, $4 beyond | 1M | ~750K | $2 |
| Grok 4.3 xAI max output not published by xAI | 1M | ~750K | $1.25 |
| DeepSeek V4 Flash DeepSeek | 1M | ~750K | $0.14 |
| Claude Haiku 4.5 Anthropic | 200K | ~150K | $1 |
| o3 OpenAI | 200K | ~150K | $2 |
| GPT-4o OpenAI | 128K | ~96K | $2.5 |
Words-that-fit uses the ~1.3 tokens-per-word average for English prose; code and other languages vary. API prices are the providers' published input rates.
Context window questions, answered.
What is a context window?
The context window is the total amount of text a model can consider at once, measured in tokens: your conversation so far, any documents you include, and the reply it is writing. Anything beyond the window is invisible to the model, no matter how important it was.
Which AI model has the biggest context window?
Among the frontier chat models, the current leaders sit at about a million tokens: GPT-5.5, GPT-4.1, and Claude Opus 4.8 and Sonnet 5. Gemini's top models are in the same range. The table above has the exact number for every model, re-verified against provider docs.
How big is Claude's context window?
Claude Opus 4.8 and Claude Sonnet 5 both take 1,000,000 tokens, roughly 750,000 words. Claude Haiku 4.5 takes 200,000 tokens. Anthropic publishes these limits in its model documentation, which is what this table is verified against.
Is a bigger context window always better?
No. Models attend unevenly across very long contexts, so burying the important part of your question in half a million tokens of filler usually gives worse answers than a short, deliberate context. The window is capacity, not comprehension: what you choose to put in it matters more than how much fits.
How do I control what goes into the context window?
In most chat apps you cannot: the whole conversation is resent with every message until it stops fitting. That is the problem NodePad exists to solve. Every conversation lives on a canvas where you fork tangents onto their own paths, reference exact messages with @, and merge only the threads you choose, so the model receives exactly the context you assembled.
A big window only helps if the right things are in it. NodePad puts every conversation on one canvas so you decide exactly what each model sees.
Try NodePad free