# Mixture of Experts

> A model architecture that routes each input through selected expert subnetworks instead of activating the whole model.

- Category: Models & Training
- Stage: Established
- Also called: MoE, Sparse expert model
- Canonical: https://www.joelcsaji.com/ai-dictionary/mixture-of-experts

## Plain English

Only the most relevant parts of the model wake up for a given token.

## Example

A sparse model can have many total parameters while using only a subset during inference.

## Why it matters

MoE is one way frontier labs improve capability while managing inference cost.

## Related

- [Frontier Model](https://www.joelcsaji.com/ai-dictionary/frontier-model)
- [Model Routing](https://www.joelcsaji.com/ai-dictionary/model-routing)
- [Latency](https://www.joelcsaji.com/ai-dictionary/latency)

## Mentioned in

- [Fable 5, Export Controls, and GLM 5.2](https://www.joelcsaji.com/blog/fable-5-export-controls-and-glm-5-2)

## Sources

- [Stanford HAI: 2026 AI Index](https://hai.stanford.edu/ai-index/2026-ai-index-report)