5.4 Describe How Token Usage and Context‑window Management Affect Prototyping Cost, Latency, and Output Quality AI-TECHNICAL-PRACTITIONER Practice Questions (Page 5)

Questions 21–25

1. 21application · medium

   A team is prototyping a code-generation assistant. They want to minimize both cost and latency while maintaining code quality. The current prompt includes a large system message (1,000 tokens) and the entire project file (5,000 tokens). Which optimization would best achieve their goal?

   • ARemove the system message entirely and rely on the user's request.
   • BShorten the system message to the essential instructions and include only the relevant parts of the file.
   • CUse a model with a larger context window.
   • DIncrease the max tokens for the response to allow longer code.
   Reveal Solution Discussion Select an answer first
2. 22foundation · easy

   How are token counts typically calculated for text input and output in an LLM API call?

   • AThe total is calculated by multiplying the number of words by a fixed factor.
   • BThe total is the number of characters divided by four.
   • COnly the input is counted; output tokens are free.
   • DThe input and output are each counted separately, and the total is the sum of both.
   Reveal Solution Discussion Select an answer first
4. 23application · medium

   A startup is prototyping a legal-document summarizer. The team wants to keep costs low while maintaining high-quality summaries. They currently send the entire document (about 30,000 tokens) to the model on every call. Which strategy best balances cost and quality?

   • ASend the entire document but reduce the max output tokens to 100.
   • BUse a model with a larger context window to fit the whole document.
   • CPre-summarize sections of the document and then ask the model to combine those summaries.
   • DUse a sliding window that sends only the first 1,000 tokens of the document.
   Reveal Solution Discussion Select an answer first
5. 24foundation · easy

   What is a common effect on output quality when a request nears the context-window limit?

   • AThe output is always longer to fill the remaining space.
   • BThe model automatically summarizes the conversation to improve quality.
   • CThe model may lose track of earlier parts of the conversation.
   • DThe output becomes more accurate because the model has more information.
   Reveal Solution Discussion Select an answer first
6. 25foundation · easy

   What can happen if the output tokens reach the context-window limit before the response is complete?

   • AThe model continues generating by discarding earlier input tokens.
   • BThe API returns an error and the user must retry with a shorter prompt.
   • CThe model automatically summarizes the response to fit the limit.
   • DThe response is truncated, cutting off the rest of the answer.
   Reveal Solution Discussion Select an answer first