A company choosing a foundation model for Amazon Bedrock wants to know how much text or context can be included in a single prompt. Which factor determines that limit?

The context window, also known as the token limit or context length, defines the maximum number of tokens (words or sub-words) that a foundation model can process in a single input prompt and generate in its output. This directly determines how much text or context can be included in a prompt. Exceeding this limit will result in truncation of the input or an error. Temperature controls the randomness of the model's output, not the input length. Batch size refers to the number of independent prompts processed simultaneously, which is an efficiency parameter and doesn't limit individual prompt length. Model size (number of parameters) indicates the model's overall capacity and knowledge but doesn't directly specify the maximum context it can handle in one go; rather, the context window is a specific architectural parameter.