The context window is the limit of how much text an AI model can process at once in a single conversation – including chat history, system instructions, uploaded documents, and the generated response. It's measured in tokens (roughly 1 token = 0.75 words).
Example sizes: GPT-4o: 128,000 tokens (~96,000 words / ~190 A4 pages). Claude 3.5 Sonnet: 200,000 tokens (~150,000 words / ~300 pages). Gemini 1.5 Pro: up to 1,000,000 tokens.
Why does this matter for enterprise AI? If you want an agent to analyse a long contract, process the full email history with a client, or keep an entire project context in mind – you need a model with a large enough context window.
However, with longer conversations and more context, models sometimes "forget" information from the beginning – the so-called "lost in the middle" effect. This is why RAG and structured agent memory are more important than just a large context window.