Stage one
Every model answers first.
Your question goes to all council models in parallel, and each writes its own independent response.
Open source · iOS and Android
Ask a council of models, not just one.
LLM Council sends your question to several AI models at once, has them rank each other's answers without knowing who wrote what, then a chairman model synthesizes the final reply. An open-source Flutter app, inspired by Andrej Karpathy's llm-council.
MIT licensed. Built in a single day, in public.
How it works
Three models, one considered answer.
Stage two
They rank each other, blind.
Each model sees the others' answers with the names stripped off, labeled A, B, C, and ranks them without knowing who wrote what.
Stage three
A chairman writes the verdict.
One chairman model reads every answer and the rankings, then synthesizes a final reply weighted by the peer review.
Yours to keep
The whole council stays on the phone.
Conversations are stored locally, and your OpenRouter key is held in encrypted storage on the device.
Under the hood
Built to be read and run.
- Framework
- Flutter and Dart, on iOS and Android
- Models
- Several at once, through the OpenRouter API
- Method
- Parallel answers, blind peer ranking, chairman synthesis
- Storage
- Councils on-device, the API key in encrypted storage
Run it
Clone it, add an OpenRouter key, run.
git clone https://github.com/sayed3li97/llm_council_app.git
cd llm_council_app
flutter pub get
flutter runAn OpenRouter key picks the council. The full setup is in the repository README.
LLM Council is open source under the MIT license.