LONG-CONTEXT LLMS
The best long-context LLMs, ranked by window size.
Llama 4 Scout currently offers the largest context window at 10,000,000 tokens — about 7,500,000 English words in one request. The ranking below covers every model in our daily-verified catalog, with each window translated into words and pages.
Catalog verified
Ranking: context window size
| # | Model | Provider | Context (tokens) | ~Words | ~Pages | Input / 1M |
|---|---|---|---|---|---|---|
| 1 | Llama 4 Scout | Meta | 10,000,000 | 7,500,000 | 15,000 | Self-hosted |
| 2 | Gemini 3.6 Flash | 1,048,576 | 786,000 | 1,573 | $1.50 | |
| 3 | Gemini 3.5 Flash-Lite | 1,048,576 | 786,000 | 1,573 | $0.30 | |
| 4 | Gemini 3.5 Flash | 1,048,576 | 786,000 | 1,573 | $1.50 | |
| 5 | Gemini 3.1 Pro Preview | 1,048,576 | 786,000 | 1,573 | $2.00 | |
| 6 | Gemini 3.1 Flash-Lite | 1,048,576 | 786,000 | 1,573 | $0.25 | |
| 7 | Gemini 2.5 Pro | 1,048,576 | 786,000 | 1,573 | $1.25 | |
| 8 | Gemini 2.5 Flash | 1,048,576 | 786,000 | 1,573 | $0.30 | |
| 9 | Gemini 2.5 Flash-Lite | 1,048,576 | 786,000 | 1,573 | $0.10 | |
| 10 | GPT-5.6 Sol | OpenAI | 1,047,576 | 786,000 | 1,571 | $5.00 |
| 11 | GPT-5.6 Terra | OpenAI | 1,047,576 | 786,000 | 1,571 | $2.00 |
| 12 | GPT-5.6 Luna | OpenAI | 1,047,576 | 786,000 | 1,571 | $0.20 |
| 13 | GPT-5.5 | OpenAI | 1,047,576 | 786,000 | 1,571 | $5.00 |
| 14 | GPT-5.5 Pro | OpenAI | 1,047,576 | 786,000 | 1,571 | $30.00 |
| 15 | GPT-5.4 | OpenAI | 1,047,576 | 786,000 | 1,571 | $2.50 |
Methodology. Word and page estimates use English-prose averages (0.75 words per token, 500 words per page). Windows are the standard API limits from official provider documentation, verified 2026-08-11; some providers charge tiered rates above 200K prompt tokens (shown on each model page). Check whether a specific document fits with the words to tokens converter.
Keep exploring
See all windows in tokens, words, and pages on context windows compared, weigh window against price in model comparisons, find the cheapest LLM APIs, or estimate long-prompt costs with prompt caching.
FAQ
Long-context LLM FAQ
Which LLM has the largest context window?
Llama 4 Scout from Meta leads with a 10,000,000-token context window — roughly 7,500,000 English words or 15,000 pages in a single request, as of 2026-08-11.
How many words fit in a 1M-token context window?
About 750,000 English words, or roughly 1,500 pages at 500 words per page. Code, JSON, and non-English text tokenize heavier, so real documents often fit 20-40% less than the prose estimate.
Does a bigger context window cost more to use?
Yes, in two ways: you pay for every input token you send, so filling a 1M-token window costs 1M tokens of input on every request; and several providers charge a higher per-token rate above a threshold (often 200K tokens). Long context is powerful but expensive to fill on every call — prompt caching and retrieval usually beat brute-force stuffing.
Do models actually use the whole context window well?
Not uniformly. Retrieval quality tends to degrade for content buried in the middle of very long prompts (the "lost in the middle" effect), so position key instructions at the start or end, and verify with your own evals before relying on full-window recall.