FROM MODEL TO APPLICATION

CHECK YOUR UNDERSTANDING • Why is an API compared to a restaurant menu? ◦ Answer: Because it lets you order work through agreed rules without needing to know how the kitchen operates. • What are the three most common message roles? ◦ Answer: System, user, and assistant. • Why must the whole conversation be resent on every call? ◦ Answer: Because the model is stateless and keeps no memory between calls. • Name one advantage and one drawback of running an open-weight model yourself. ◦ Answer: You control your data and costs at scale, but you must manage the hardware and operations. • Why are GPUs important to language models? ◦ Answer: Because models need huge numbers of calculations done in parallel, which GPUs do far better than ordinary chips.