Model right-sizing: the fitting model, not the most expensive one

There's a reflex to reach for the biggest, newest model for every task. It's usually the wrong call — expensive, slower, and no more accurate for the job in front of you.

Right-sizing means matching the model to the task: a small model for classification and extraction, a larger one only where genuine reasoning earns its keep. The result is lower cost, faster responses, and often more reliable output because the system is simpler.

We treat model choice as an engineering decision with a budget attached — measured, not assumed.