A context window is the total amount of text a model can hold in a single request, counted in tokens, with your input and its output sharing one budget. It sets how much the model can keep in view during one exchange, and anything past the limit drops out of sight.
The one-sentence definition
A context window is the cap on how many tokens a model can process at once, shared between what you type in and what it writes back. A token is roughly a short piece of text: a few characters, a whole word, or part of one. Once a conversation grows past the cap, the earliest content gets pushed out or truncated, and the model can no longer reference what it can’t see.
An example
Paste a long report into a chat and ask for a summary, and the report, your question, and the model’s answer all have to fit inside the same window. If the whole report fits, the model reads it in one pass; if it overflows, the later sections can get cut. You can try this in Claude or Google Gemini. Keep one thing in mind: even when the material fits, details scattered across a long document can slip through. With a tool like Google Gemini, large-file analysis can miss dispersed details and needs a source review, and with Claude the web findings and generated conclusions still deserve a check against the primary source.
How it differs from related ideas
A context window is not the same thing as the model’s memory, and the two are worth keeping apart. Training data is what the model learned in advance and holds fixed in its weights; the context window is whatever you load into the current conversation, and it clears once the request ends. Some products add a separate long-term memory or a retrieval step to fill the gap, but that runs alongside the window rather than making the window itself bigger.
When it matters
The window size matters most when you’re working with long documents, extended multi-turn chats, or large blocks of code, because it decides whether the model can take everything in at once. A bigger window lets you feed in more at a time, though larger windows often come tied to higher plans: with Claude, usage allowances and advanced features depend on the plan, so confirm your quota before you rely on it. And filling the window to the brim doesn’t mean the model weighs every section equally; details sitting in the middle are the easiest to overlook, so check important conclusions back against the source.
