A team uses Opus for 100% of calls. They identify that 60% of calls ar — CCA Advanced Practice — Claude Cert Academy

A team uses Opus for 100% of calls. They identify that 60% of calls are simple classification. What single optimization has the highest cost impact?

• Reduce output max_tokens
• Route the 60% classification segment to Haiku
• Add streaming
• Switch all calls to Batch API

Answer: Route the 60% classification segment to Haiku

Routing 60% of traffic from Opus (~$15/MTok) to Haiku (~$0.80/MTok) is roughly a 19× cost reduction on that segment, dropping overall cost by ~50%. Output token limits and streaming don't change cost; Batch API only works for non-interactive paths.

Continue to Claude Cert Academy