Petals

by BigScience workshop · MIT · 10,697 GitHub stars · Quiet · v2.2.0, Sep 2023

The original 'BitTorrent for AI': volunteers each hosted a few layers of a big model over the internet. No work since 2024, and the public swarm no longer answers.

Mac: Partly
NVIDIA PC: Yes
AMD PC: Partly
Linux, processor only: No
Raspberry Pi and boards: No
Windows: Partly
Android: No
iPhone, iPad: No

Go to Petals Source code

At a glance

Best for: Historical interest, or a private research swarm of trusted GPU servers.
Runs on: Linux NVIDIA servers; Macs and Windows only partly.
How it splits a model: Pipeline across volunteers.
Connections: Built for the internet.
How devices join: Automatic (a DHT).
Models: Hugging Face models.
API: Python only.
Rules: On the public swarm, strangers' machines see your prompts.

Get started

Linux

pip install git+https://github.com/bigscience-workshop/petals

Limits to know

No commits since August 2024; its health page timed out on 2026-11-06.

We don’t estimate speeds for Petals.

Sources:

All clustering tools

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.