Option A correctly matches the model tier to the workload. An interactive, high-volume workflow with light reasoning places greater importance on response speed, operational efficiency, and cost than on maximum reasoning depth. Haiku is positioned as the fast, lightweight Claude model for responsive tasks, making it the appropriate starting choice.

Option B uses a criterion unrelated to performance. Option C is incorrect because model tiers differ in capability, latency, and cost characteristics. Option D prioritizes advanced reasoning even though the scenario explicitly states that reasoning demands are light and rapid responses are required. Selecting a more capable model than the task needs can increase latency and expense without producing a meaningful business benefit.

Model selection should still be validated empirically. The team should define an acceptable quality threshold, test representative customer-success requests, measure response latency and task accuracy, and confirm that Haiku meets the required service level. If the real workload contains complex exceptions, the workflow could route only those cases to a more capable model. Anthropic’s current learning materials characterize Haiku as its fastest option for lightweight work, while its prompt-engineering guidance identifies model selection as a direct lever for improving cost and latency. Anthropic’s Claude learning resources

================