"What Is a Model Context Window? (2026 Plain-English Explainers)"
The context window is the most important spec most people ignore. This explainer covers what it is and why bigger is a big deal.
What it is
A context window is the maximum text a model can hold in one go — your prompt plus its reply, measured in tokens (roughly 4 characters each). Once it fills, the model forgets the start.
How big in 2026
- Claude offers up to 1 million tokens — a full book or mid-size codebase.
- Gemini reaches 2 million tokens — a small library.
- GPT-class models sit in the 1M-class range.
See the Claude and Gemini reviews for how each uses it.
What it unlocks
- Paste an entire codebase and ask cross-file questions.
- Analyze a 300-page PDF in one prompt.
- Keep long conversations without losing the thread.
The catch
Bigger windows cost more and can slow replies; “more context” is not automatically “better answers.” Retrieve the right chunks (see RAG) rather than dumping everything.
FAQ
How many words is 1M tokens? Roughly 750,000 words — about 7 novels.
Does a bigger window mean smarter? Not directly; it means more can be held at once. Quality still depends on the model.
Do I pay per token? Yes — both input and output tokens are metered, so huge windows can raise cost.
Bottom line
The context window is the model’s working memory. In 2026, 1M–2M tokens let you feed whole books and codebases — but retrieve smartly to keep answers sharp.