LLM context windows, ranked

Llama 4 Scout (17B-16E Instruct) (Meta) has the largest context window of the generally-available models we track, at 10M tokens. GPT-5.6 Sol is next at 1.1M tokens. Context windows are read from official provider documentation and refreshed daily.

A context window is the maximum number of tokens — roughly ¾ of a word each — that a model can consider at once, spanning your prompt, any retrieved documents or code, the conversation history and the model’s own reply. A larger window lets you feed in more source material (whole codebases, long PDFs, long chats) in a single request without chunking.

No. Effective recall often degrades well before the stated limit — a model with a 1M-token window may still miss details buried in the middle of a very long prompt. Bigger windows also cost more, because you pay per input token. Match the window to what you actually need to fit, then verify recall on your own long-context task.

32 of the generally-available models we track offer a context window of at least 1M tokens, and the cheapest of them is Qwen-Flash (Alibaba) at a blended $0.138 per 1M tokens. Sort the table by context to see them ranked, or by blended price to find the cheapest large-context option.