Docs Project

Contribute

Ways to help BOAR: benchmark models on your phone, find the questions it gets wrong, build knowledge packs, and send code to the open-source Android app.

BOAR is for anyone curious about what a phone can really do. You don't need to write code to help.

Test a model before your next trip

Found a promising GGUF model on Hugging Face? Add it from the app's model browser and benchmark it on your phone. You get load time, time to the first word, speed, peak memory and every answer side by side.

Share your results

Different phones, different chips, different numbers. A result from your device is data nobody else has. Post it in the Discord or in an issue with the .jsonl file.

Break it

Ask the questions a small model gets wrong. Find where search pulls in nonsense. Open an issue with the question, the answer and the sources it showed.

Build a knowledge pack

Make a pack for a place, a trip or a subject from a list of Wikipedia titles, and share it. See Knowledge packs.

Build the missing pieces

The bigger jobs, each one a step towards the model we're waiting for:

  • Expert-aware caching for mixture-of-experts models that stream from storage.
  • Bigger offline knowledge packs, and a shared library of them.
  • Better routing between small and large models.
  • Offline voice with a speech model that runs on the phone.
  • iOS.

Start with Build from source and the open issues. The code is MIT.

Talk to us

Edit this page on GitHub · Something wrong? Open an issue