Long Context

What is long context?

"Context" is everything an AI can "see" while answering — your prompt, the conversation so far, the documents you feed it, all added together. A normal model's context is a sticky note that fills up after a few paragraphs. A "long context" model is a notebook that can hold an entire novel: hundreds of thousands of words, hundreds of files, at once.

What's it good for?

Feed it a whole book
To analyze a book, you used to have to chop it up and stitch it back together — fiddly and lossy. Long context lets you drop the whole thing in.
Long conversations stop forgetting
The longer the context, the more the AI remembers from way back, so multi-turn chats stop contradicting themselves.
Digest oceans of material
Dozens of contracts, an entire code repo, a year of logs — load them all and let the AI analyze globally and find links.

Why do enterprises care so much?

One survey found 78% of enterprise buyers rank long context their top requirement when picking a model. Real work sprawls across huge amounts of documents and background — if the context is too short, the AI literally can't hold your whole business.

What's the catch?

Long context isn't free: longer inputs mean more compute, slower responses, and "fits in" isn't the same as "actually remembers" — attention can sag in the middle of a long text. So beyond "long enough," "really reads it all" is the real test.

Bottom line: long context gives an AI a bigger working memory, so it can hold a whole book and remember a whole conversation.