model-selection

Route tasks by complexity: assign Opus to deep architecture and trade-off decisions, Sonnet to throughput-oriented routine implementation, and Haiku to fast, cheap subagent lookups and summaries to balance reasoning quality and cost.

The plan-execute-review workflow uses Opus to plan the architecture, Sonnet to execute the high-throughput implementation, and Opus again to review the output, combining strong reasoning with efficient cost-aware execution.

Yes, you can assign subagents different models based on whether they need reasoning or only execution and search, routing Haiku to fast lookup tasks and reserving Opus for subagents handling deep trade-off analysis.

Use fast mode for throughput-focused output when the task requires high-volume routine implementation rather than deep architectural reasoning, allowing Sonnet to handle the execution efficiently at a lower cost.

Model selection applies directly to .NET development tasks across workflow planning, implementation, debugging, and PR review, matching Opus, Sonnet, and Haiku to the appropriate routine and complex scenarios.

Using a high-reasoning model like Opus for routine implementation or a fast model like Haiku for complex architecture creates slow, expensive, or low-quality responses, making proper task-to-model routing essential.