# Mesh-LLM/mesh-llm

Distributed AI/LLM for the people. Share compute privately or publicly to power your agents and chat.

Repository: https://github.com/Mesh-LLM/mesh-llm
Canonical: https://ross.abutalabs.com/products/mesh-llm
Homepage: http://meshllm.cloud/
Language: Rust
License: Apache-2.0
License Family: permissive
Topics: agents, ai, decentralized, distributed, llm
Last push: 2026-08-26T22:37:28+00:00

## Health v2 (maintenance only)
Score: 77/100 (v2, computed 2026-09-03T02:39:23.370411+00:00)
- activity 99, release rhythm 85, longevity 14
- inputs: {"age_days": 203, "days_push": 7, "days_rel": 23, "gap_med": 0.0, "n_releases_24m": 105}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3305, forks 397 (observed 2026-08-28T04:07:55.788245+00:00)

## What it is
Mesh LLM is a Rust-based distributed LLM inference runtime that pools GPUs and memory across machines into a single OpenAI-compatible API. It supports public and private mesh networking, automatic routing, and splitting large models into layer slices across nodes.

## Use cases
- run large LLMs across multiple machines without buying bigger GPUs
- pool GPUs from homelab nodes into one inference endpoint
- share compute with a public mesh to power agents and chat
- serve an OpenAI-compatible API for coding agents and existing clients
- split a model too large for one machine across a LAN cluster
- route requests between multiple locally hosted models

## When to choose
- you have several machines with modest GPUs and want to run models bigger than any single one
- you want a drop-in OpenAI-compatible endpoint backed by your own hardware
- you want to share or borrow compute via public or invite-based meshes
- you need tool calling, streaming, and structured outputs from a self-hosted runtime

## When to avoid
- you need a single-GPU setup with maximum inference throughput and no networking overhead
- you require enterprise-grade SLAs or managed hosting
- you depend on non-OpenAI-specific serving features like tensor-parallel vLLM deployments
- you cannot accept the latency of layer-split pipelines across slow network links

## Facets
- artifact type: application
- maturity: active
- function: llm-inference, http-server, api-framework, cli, agent-framework, chatbot, gpu-computing, plugin-system
- domain: large-language-models, artificial-intelligence, microservices, self-hosted, developer-tools
- platform: windows, cross-platform, cli, self-hosted, rust
- tags: distributed-inference, openai-compatible-api, p2p, model-sharding, homelab, mesh-networking, gguf, ai-agents, linux, macos

## Member repositories
- Mesh-LLM/mesh-llm (main) score 77

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:55.788245+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 2026-08-29T18:42:38.927949+00:00, confidence not recorded.
  - readme: https://github.com/Mesh-LLM/mesh-llm (fetched 2026-08-28T04:07:55.788245+00:00, sha 02231f760dc0)
  - homepage: http://meshllm.cloud/ (fetched 2026-08-29T09:35:16.926420+00:00, sha ac4c938c90db)
  - site_page: http://meshllm.cloud/docs (fetched 2026-08-29T09:35:17.309798+00:00, sha 1c3409352e9e)
  - site_page: http://meshllm.cloud/docs/pages/installing-mesh (fetched 2026-08-29T09:35:17.312932+00:00, sha 3ba1aea30194)
- Data as of 2026-08-30T08:39:29.467469+00:00.