LMForge

v0.1.7

Run AI Models on Your Own Machine

Run AI models privately on your own hardware.

Download a model and within minutes you have a fast, OpenAI-compatible AI endpoint running entirely on your machine. LMForge detects your GPU, VRAM, and drivers, installs the best inference engine automatically, and keeps your models served by an always-on daemon — closing the app never stops them.

GitHub ↓ Latest release

Demo video coming soon

🔒 Private — nothing leaves your device

⚡ One-command install, auto engine selection

🔌 OpenAI + Ollama compatible API

🧠 Multi-model orchestration

Version v0.1.7License MITCurated models 130+Platforms macOS · Linux · Windows

// install

Get running in one command

1. Install core (daemon + CLI) bash

curl -fsSL https://github.com/phoenixtb/lmforge/releases/latest/download/install-core.sh | bash

2. Install the desktop app (optional) bash

curl -fsSL https://github.com/phoenixtb/lmforge/releases/latest/download/install-ui.sh | bash

3. Pull a model and chat bash

lmforge pull qwen3:8b:4bit
lmforge run qwen3:8b:4bit

The installer starts the engine and registers it to start at login (launchd on macOS, systemd --user on Linux). The API lives at http://localhost:11430; everything LMForge owns sits under one folder — ~/.lmforge.

// why local

Cloud AI APIs vs LMForge

Same developer experience as a cloud endpoint — without the meter running or your data leaving the building.

	Cloud AI APIs	LMForge (local)
Data privacy	Prompts leave your network	Everything stays on your machine
Cost	Pay per token, forever	Pay once for hardware; runs free
Offline	Needs internet	Works fully offline
Vendor lock-in	Tied to one provider	Open models, swap any time
Compliance	Third-party data processing	Full control, on-premises
Setup	API keys, billing	One command, done