A startup needs to classify thousands of customer support emails per hour at minimal cost and fast latency. Which Claude model is MOST appropriate?

Claude Haiku is specifically designed for high-throughput, cost-sensitive, low-latency use cases. Email classification is a well-defined, relatively simple NLP task where Haiku's capabilities are more than sufficient. Using Opus for this would be like using a chainsaw to slice bread — overkill and expensive. Option A (Opus) is the most capable model but far more expensive and slower — poor fit for high-volume simple tasks. Option C is wrong — multilingual support is not uniquely associated with Sonnet. Option D is wrong — model choice has direct, significant impact on both cost and latency.