Local LLM
Agent
Agent! now supports macOS 14.6 or later, Apple Silicon and Intel, latest release.
★ 642 starsNOASSERTION
View guide →A fast LLM inference server for Apple Silicon, managed from the macOS menu bar
omlx is an open-source local llm project by jundot (jundot/omlx on GitHub). It has 22,638 GitHub stars and 1,985 forks, is released under the Apache-2.0 license, and was last updated on GitHub on 8 October 2026.
You need an Apple Silicon Mac; pick the models that fit your unified memory.
Install the macOS app or the Homebrew package, or clone the repo for a source install.
Install with Homebrew, start the background service, and optionally add MCP support.
Point oMLX at your model directory, load a model, and query the OpenAI-compatible API.
Batches concurrent requests so multiple users or agents can share one Mac server efficiently.
Uses SSD caching alongside a tiered hot-and-cold KV cache to stretch Apple Silicon memory.
Start, stop and manage the server from the macOS menu bar or the brew services integration.
Serve several models at once, each with its own per-model settings.
Includes an admin dashboard and a built-in chat interface for quick testing.
Supports vision-language models and optional MCP (Model Context Protocol) integration.
Get the source code from GitHub and enter the project directory.
git clone https://github.com/jundot/omlx.git
cd omlxUse the project's Homebrew tap, then start oMLX as a background service.
brew tap jundot/omlx https://github.com/jundot/omlx
brew install jundot/omlx/omlx
omlx startRun the server attached to your terminal against a local model directory.
omlx serve --model-dir ~/modelsIt is maintained by jundot and contributors on GitHub.
Yes - oMLX is Apache-2.0 licensed and free to use.
No. It runs open models locally on your Mac, so there are no API keys or per-token costs.
Compared with Ollama, oMLX is MLX-native and macOS-focused, with continuous batching, SSD/tiered KV caching and menu-bar management. Compared with LM Studio, it is fully open source and server-oriented.
Install the macOS app or Homebrew package, run omlx start, load a model, and send a request to the OpenAI-compatible endpoint.
Agent! now supports macOS 14.6 or later, Apple Silicon and Intel, latest release.
Run siliconscope locally: sudoless apple silicon system monitor (native swiftui g.
Run a 70B LLM on a single 4GB GPU — layer-by-layer inference with almost no memory tricks skipped
Alibaba's lightweight, blazing-fast inference engine for on-device LLMs and edge AI.
Open bilingual (Chinese-English) 6B chat LLM with 32K-context variant and low-memory INT4 inference.
TokenSpeed — a speed-of-light LLM inference engine designed for agentic workloads, with TensorRT-LLM-level performance and vLLM-level usability
The most-starred other Local LLM projects in the LibreAI catalogue — the realistic shortlist if omlx is not the right fit.
The simplest way to run open language models on your own machine.
User-friendly AI Interface (Supports Ollama, OpenAI API, ...)
LLM inference in C/C++
A high-throughput and memory-efficient inference and serving engine for LLMs
🤯 LobeHub is your Chief Agent Operator, organizing your agents into 7×24 operations by hiring, scheduling, and reporting on your entire AI team.
Free and open — grab the source, star the repo, and start creating.