Context length is the maximum number of tokens an LLM can process in a single call (input + output combined). It determines how much text you can send and receive. Exceeding it causes errors or truncation and longer contexts increase cost and latency.