A question of fit
Model size is only one part of a system. Relevant data, clean inputs and a clear task can determine whether a result is useful.
What good data looks like
Good data has known provenance, sensible permissions and a level of freshness appropriate to the task. It includes examples that reflect the work you actually need to do.
Avoid simple comparisons
A larger model can be useful for some tasks. The practical decision is to compare systems on the same real examples and judge quality, cost and reliability together.
Key takeaways
Choose a system using your task and your evidence. Size alone is not a useful purchasing or design rule.