# b4rtaz/distributed-llama

Distributed LLM inference. Connect home devices into a powerful cluster to accelerate LLM inference. More devices means faster inference.

Repository: https://github.com/b4rtaz/distributed-llama
Canonical: https://ross.abutalabs.com/products/distributed-llama
Language: C++
License: MIT
License Family: permissive
Topics: distributed-computing, llama2, llm, llm-inference, neural-network, llms, open-llm, distributed-llm, llama3
Last push: 2026-07-05T16:47:20+00:00

## Health v2 (maintenance only)
Score: 79/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 91, release rhythm 68, longevity 71
- inputs: {"age_days": 1003, "days_push": 59, "days_rel": 212, "gap_med": 4, "n_releases_24m": 40}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3045, forks 246 (observed 2026-08-28T04:07:39.490990+00:00)

## What it is
A C++ application that connects multiple home devices into a cluster for distributed LLM inference using tensor parallelism. It supports Llama, DeepSeek, and Qwen models on CPU (ARM and x86_64 AVX2) and experimentally on GPU via Vulkan.

## Use cases
- run llama 3 70b across multiple mac minis
- speed up llm inference by pooling home devices
- serve a large language model on raspberry pi cluster
- distributed inference of qwen 3 models
- run deepseek r1 distilled model locally on weak hardware
- tensor parallel llm inference over ethernet

## When to choose
- you have several networked devices and want to run models too large for any single one
- you want CPU-only distributed inference without a GPU
- you want a simple single-command setup for popular open LLMs

## When to avoid
- you have a single powerful GPU that fits the model
- you need production-grade serving with APIs and batching
- you need broad model format support beyond supported families

## Facets
- artifact type: application
- maturity: active
- function: llm-inference, cli, rpc
- domain: large-language-models, microservices, artificial-intelligence
- platform: windows, cpp, cli
- tags: tensor-parallelism, llama, qwen, vulkan, raspberry-pi, cluster-computing, cpu-inference, linux, macos, gpu

## Member repositories
- b4rtaz/distributed-llama (main) score 79

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:39.490990+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 2026-08-30T07:28:57.578954+00:00, confidence not recorded.
  - readme: https://github.com/b4rtaz/distributed-llama (fetched 2026-08-28T04:07:39.490990+00:00, sha cbd1fcc720e7)
- Data as of 2026-08-30T08:39:29.467469+00:00.