A systematic decision framework for selecting between Claude 3.5 Haiku, Claude 3.5 Sonnet, and Claude 3 Opus across four dimensions: task complexity, latency sensitivity, cost at scale, and output quality. Includes detailed pricing comparisons (Opus costs ~19x more than Haiku), concrete use-case routing recommendations, Python code examples for dynamic model selection, cost projections at 1M requests/month, and guidance on when to upgrade or downgrade tiers. Also covers a cascade pattern for hybrid multi-tier architectures to optimize cost without sacrificing quality.
Table of contents
Table of ContentsWhy Model Selection Matters More Than Prompt EngineeringClaude Model Lineup at a GlanceThe Decision Framework: Matching Tasks to ModelsTask-Based Recommendations: Concrete Use CasesImplementing Model Selection in the Anthropic APICost/Performance Analysis: Running the NumbersTier Upgrade and Downgrade GuidanceBuild Your Own Selection PolicySort: