Large Model Reasoning Test

Different reasoning tasks need different routing choices.

Tasks: mathematical reasoning, logical reasoning, and code reasoning.

Finding: one candidate performed better on code tasks, while another handled structured logic more consistently.

Conclusion: route by task type instead of using one default model for everything.