---
title: "Mixture-of-Agents (MoA) · agent orchestration · Agent Fieldbook"
url: https://agentfieldbook.org/design/orchestration/moa/
description: "Layers of LLMs where each layer's agents take all prior-layer responses as auxiliary context, then a final aggregator synthesises. Quality compounds across layers."
section: "Design · Orchestration"
source: Agent Fieldbook — generated from the published page
---

# Mixture-of-Agents (MoA)

Layers of LLMs where each layer's agents take all prior-layer responses as auxiliary context, then a final aggregator synthesises. Quality compounds across layers.

Verified against [arxiv.org](https://arxiv.org/abs/2406.04692)

- [arxiv.org](https://arxiv.org/abs/2406.04692)

## How it works

## At a glance

| Field | Value |
| --- | --- |
| Technique | Ensemble |
| Cost lever | Higher, runs many models per query |
| Latency lever | Higher, sequential layers |
| Quality lever | Open-weight ensemble beats GPT-4o on AlpacaEval 2.0 |
| Example | Together AI MoA· Self-MoA |

## When to use

Quality-critical tasks where you can trade cost/latency for a better answer, and want to avoid a single frontier dependency.

## Basis

## Verified against

- [https://arxiv.org/abs/2406.04692](https://arxiv.org/abs/2406.04692)

The corpus does not rewrite vendor documentation. Where copy is quoted verbatim it is marked as such and attributed to its source.