What is Mixture of Experts (MoE)?

An architecture that activates only part of the model per request, giving large-model quality at lower compute cost.

This is one of 36 terms in the Benchquill AI glossary. Knowing it helps when you compare AI tools and AI models on price and capability.