Context Window
How much text an AI can see and remember at once.
Every LLM has a context window — the maximum number of tokens it can process in a single interaction. Everything the model can "see" — your system prompt, conversation history, documents you paste in — must fit within this limit.
If the window is 128,000 tokens (~96,000 words) and your conversation exceeds that, earlier messages are dropped. The model has no memory of them.
Modern frontier models: GPT-5.5 supports 1M tokens, Claude Opus 4.7 and Sonnet 4.6 also support 1M, and Gemini 3.1 Pro extends to 2M — the largest of any frontier model as of mid-2026. This is critical for analysing long documents, maintaining long conversations, or processing entire codebases.
In plain terms
The context window is like a whiteboard in the room. The AI can only reference what's written on it — once it fills up, you must erase something to write more.