A company is making 1 million Claude API calls per month with a 500-token system prompt. What would have the BIGGEST impact on reducing their monthly API costs?
Select an answer to reveal the explanation.
Short Explanation and Infographic
A 500-token system prompt × 1 million calls = 500 million tokens per month in repeated system prompt costs. Caching it converts almost all of those to cheap cache reads — massive savings.
Full explanation below image
Full Explanation
Prompt caching's impact is most dramatic when: the same content is repeated across many calls, and that content is substantial. A 500-token system prompt repeated 1M times = 500M input tokens at full price. With caching, the first call creates the cache entry (full price), and subsequent ~999,999 calls read from cache at reduced price (typically 10-30% of input token price), saving potentially 70-90% of system prompt token costs. Option A (HTTP/S) doesn't affect pricing — it's about transport security, not costs. Option C (stop_sequences) has no pricing impact. Option D (max_tokens reduction) could save output tokens but only if responses are currently being truncated at the limit.