A technique where a small model (the "student") is trained to replicate the behavior of a large model (the "teacher"), producing a faster and cheaper model to operate with controlled quality loss.
In AIF-C01 (Domain 3) it sits at the highest-effort end of the customization scale, alongside pre-training: useful when you need to reduce inference latency/cost while retaining most of the large model's capability.
Model distillation is a concept that can show up on the exam. Check the full guide to see it in context.