Quick answer
The context window is the maximum amount of text an AI can "see" at once — including your question, all previous messages in the conversation, and any documents you have attached. Bigger context windows = AI can handle longer documents and longer conversations without forgetting earlier parts. In 2026, context windows range from 32k tokens (~24k words) up to 10 million tokens (Gemini).
When people talk about AI specs, "context window" comes up constantly. It is one of the most-discussed and most-confused numbers in AI. Here is what it actually means and why it matters.
How does the context window work?
Every time you send a message to an AI, the entire conversation history plus any attached documents get bundled into a single input. That input has a maximum length — the context window. If the conversation goes beyond it, the oldest messages start getting cut off, and the AI "forgets" the beginning.
How big are context windows in 2026?
- GPT-5: 200,000 tokens (~150,000 words — a full novel)
- Claude Opus 4.7: 500,000 tokens (~375,000 words)
- Gemini Ultra 2.0: up to 2 million tokens (~1.5 million words)
- Gemini 1.5 Pro experimental: 10 million tokens (over 7 million words)
- Open source Llama 4: up to 128,000 tokens
Why do bigger context windows matter?
Bigger context windows let you do things impossible with smaller ones — analyse an entire codebase, summarise a full book, work through a long contract, or carry on a conversation for hours without the AI forgetting where you started. In 2023, you could fit a few pages. In 2026, you can fit a small library.
A million tokens of context is roughly: 1500 pages of text, OR 50 hours of speech transcribed, OR a 500-page novel and all its reviews combined. That is the scale modern AI can hold in working memory.
Are bigger context windows always better?
Not quite. Bigger windows cost more (you pay per token), and AI accuracy can drop on very long contexts ("lost in the middle" problem). Most everyday tasks fit in 32k tokens comfortably. Use the bigger windows only when you actually need them.
Related reading
Bottom line
Context window = AI memory size. Bigger windows = more you can pack in. For everyday use, you rarely notice. For document analysis, codebases, and long conversations, it is one of the most important specs.
