# bigscience-workshop/petals

🌸 Run LLMs at home, BitTorrent-style. Fine-tuning and inference up to 10x faster than offloading

Repository: https://github.com/bigscience-workshop/petals
Canonical: https://ross.abutalabs.com/products/petals
Homepage: https://petals.dev
Language: Python
License: MIT
License Family: permissive
Topics: bloom, deep-learning, distributed-systems, language-models, large-language-models, machine-learning, neural-networks, pytorch, volunteer-computing, pipeline-parallelism, tensor-parallelism, guanaco, llama, chatbot, gpt, transformer, nlp, pretrained-models, falcon, mixtral
Last push: 2024-09-07T11:54:28+00:00

## Health v2 (maintenance only)
Score: 23/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 8, longevity 100
- inputs: {"age_days": 1544, "days_push": 725, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 10521, forks 642 (observed 2026-08-28T04:10:42.212691+00:00)

## What it is
Petals is a Python library that lets you run and fine-tune large language models (Llama 3.1, Mixtral, Falcon, BLOOM) on a BitTorrent-style distributed network of volunteer GPUs. It integrates with Hugging Face Transformers so models load and generate as if they were local, while layers are served across a community swarm.

## Use cases
- run 100B+ parameter LLMs without a high-end GPU
- generate text with Llama 405B from a laptop or Google Colab
- fine-tune large language models on consumer hardware
- build a chatbot backed by a distributed LLM
- contribute idle GPU capacity to a community model-serving swarm
- experiment with model hidden states and custom inference paths

## When to choose
- you want to run or fine-tune very large models without expensive hardware
- you need PyTorch/Transformers flexibility rather than a fixed hosted API
- you want to donate GPU resources to a public inference network
- you're doing research on distributed inference or model parallelism

## When to avoid
- you need guaranteed latency or throughput for production workloads
- you handle sensitive data that can't leave your machine (unless you set up a private swarm)
- you need a fully self-contained offline deployment
- you require strict SLAs or enterprise support

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, llm-training, machine-learning, deep-learning, nlp
- domain: large-language-models, machine-learning, deep-learning, microservices
- platform: python
- tags: distributed-inference, volunteer-computing, bittorrent-style, transformers, pytorch, fine-tuning, swarm, bloom, llama, mixtral, falcon, natural-language-processing, linux, macos, gpu, docker

## Member repositories
- bigscience-workshop/petals (main) score 23

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:42.212691+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:18:34.758735+00:00, confidence not recorded.
  - readme: https://github.com/bigscience-workshop/petals (fetched 2026-08-28T04:10:42.212691+00:00, sha 330e9aca2042)
  - homepage: https://petals.dev (fetched 2026-08-29T08:17:54.804270+00:00, sha 6feb3fe77d1f)
  - registry_pypi: https://pypi.org/pypi/petals/json (fetched 2026-08-29T08:17:54.806761+00:00, sha ed64fe5bb5a0)
- Data as of 2026-08-30T08:39:29.467469+00:00.