Which models do you use?

“Whichever fits the task, the latency budget and the cost ceiling. We benchmark candidates against your own evaluation set rather than a public leaderboard, and we keep the model layer swappable so a better or cheaper model can be adopted later without a rebuild.”