What is Model Distillation?
Distillation trains a smaller, cheaper "student" model to mimic a larger "teacher" model, keeping most of the quality at a fraction of the cost and latency. It is a common way to make AI affordable in production.
Distillation lets teams deploy fast, low-cost models for high-volume tasks while reserving larger models for the hardest cases. It is part of a broader toolkit — alongside quantisation and caching — for controlling inference cost at scale.