LLM Council: AI Queries with Multi-Model Debates

In AI’s dynamic landscape, the LLM Council—crafted by Andrej Karpathy—transforms solo model chats into collaborative powerhouses. This open-source tool on GitHub queries multiple LLMs via OpenRouter, sparking anonymous debates for unbiased, superior answers. Ideal for coders, researchers, and AI enthusiasts exploring multi-agent systems.

LLM Council Essentials

The LLM Council simulates an expert panel:

• Query Broadcast: Hits models like GPT-4o and Claude for initial takes.
• Blind Reviews: Peers rank outputs anonymously on accuracy and insight.
• Chairman Wrap-Up: Synthesizes into one refined response.

Built with FastAPI and React, it’s fast, private, and customizable—boosting answer quality by 20-30% in tests.

Quick Setup for LLM Council

1. Clone repo, run uv sync (Python) and npm install (frontend).
2. Add OpenRouter API key to .env.
3. Launch via ./start.sh; access at localhost:5173.

Edit config.py for your model mix. Dive deeper into tools like this in our awesome AI apps, LLM applications, and AI agents roundup.

Why Choose LLM Council?

It cuts through single-model biases, delivering consensus-driven wisdom for tough problems. Try it now—fork the GitHub repo and query away. What’s your first LLM Council challenge? Comment below!