Using multiple models balances cost, speed, and accuracy. You get better...

https://wiki-coast.win/index.php/The_AI_Verification_Problem:_Why_Your_Models_Are_Lying_to_You_in_Harmony

Using multiple models balances cost, speed, and accuracy. You get better coverage when models fail differently. Be careful though. Aggregating results often hides dissent and risks data leaks