title: "Mixture of Agents (MoA)" category: 09-agentic-patterns level: advanced stability: experimental description: "Apply mixture of agents (moa) in AI agent workflows." added: "2025-03"
Mixture of Agents (MoA)
Category: agentic-patterns
Skill Level: advanced
Stability: experimental
Added: 2025-03
Description
Multiple LLM agents independently generate responses; an aggregator model combines their outputs into a final, higher-quality answer. Exploits the diversity of different model families.
Example
Query: "Explain quantum entanglement simply"
Agent 1 (GPT-4o): "Imagine two coins always landing opposite..."
Agent 2 (Claude 3): "Entanglement links particles so measuring one..."
Agent 3 (Gemini): "Like a magic pair of dice that always match..."
Aggregator: synthesizes all three → best combined explanation