A curated list of awesome projects, tools, models, and resources for Apple MLX — the ML framework for Apple Silicon.
Run AI models on your Mac at full speed. This list covers everything built on MLX — inference servers, training tools, audio/speech/vision models, Swift packages, and more. Whether you want a ChatGPT-like experience offline or you're building an AI app, start here.
MLX is Apple's open-source ML framework designed for Apple Silicon. If you have an M1/M2/M3/M4 Mac:
- Use all your RAM for models — Your Mac's memory is shared between CPU and GPU. A 32GB Mac can run a 30GB model — no copying between chips.
- Faster than CUDA workarounds — Native Metal GPU acceleration, not a compatibility layer.
- Growing fast — 120+ projects, 2000+ optimized models on HuggingFace, and active community.
"I just want to chat with a local AI on my Mac" → Install Klee (native app, one click) or chat-with-mlx (web UI)
"I want an OpenAI-compatible local API server"
→ Install Rapid-MLX (pip install vllm-mlx && rapid-mlx serve qwen3.5-9b)
"I want to fine-tune a model on my Mac" → Use mlx-tune (SFT, DPO, GRPO) or Apple's built-in mlx-lm LoRA
"I'm building a Swift/iOS app with on-device AI" → Start with mlx-swift-lm + fullmoon-ios as reference
Submit your project → PRs and issues welcome! Language legend: 🐍 Python · 🦅 Swift · 🟨 JS/TS · 🦀 Rust · 🐹 Go
- Core Framework
- Inference & Serving
- Training & Fine-tuning
- Audio & Speech
- Image & Video Generation
- Vision & Multimodal
- Embeddings & RAG
- Swift Ecosystem
- Benchmarks
- Apps & Demos
- Other Tools
- Models
- Learning Resources
- 🐍 mlx — Apple's array framework for ML on Apple silicon. The foundation.
- 🐍 mlx-examples — Official examples: LLMs, LoRA, Stable Diffusion, Whisper, and more.
- 🐍 mlx-lm — Apple's official LLM runner. Text generation, quantization, LoRA.
- 🦅 mlx-swift — Swift API for MLX.
- 🦅 mlx-swift-examples — Examples using MLX Swift: LLMs, image generation, training.
- 🦅 mlx-swift-lm — LLMs and VLMs with MLX Swift.
- 🟨 node-mlx — Machine learning framework for Node.js, built on MLX.
- 🐍 mlx-data — Apple's efficient framework-agnostic data loading.
- 🐍 mlx-c — Official C API for MLX.
- 🦀 mlx-rs — Unofficial Rust bindings to Apple's MLX framework.
- 🐍 mlx-graphs — Graph Neural Network library for Apple Silicon.
- emlx — MLX backend for Elixir Nx.
- ocaml-mlx — OCaml bindings for MLX with JSX syntax.
Which one should I use? Quick guide:
- Just want a GUI app? → omlx (menu bar) or LM Studio (desktop app)
- Need an OpenAI-compatible API? → Rapid-MLX (fastest, tool calling) or mlx-omni-server
- Building a Swift/iOS app? → swama or PicoMLXServer
- 🐍 omlx — LLM inference server with continuous batching & SSD caching, runs from macOS menu bar.
- 🐍 lmstudio mlx-engine — LM Studio's Apple MLX engine.
- 🐍 mlx-omni-server — Local inference server with OpenAI-compatible API for Apple Silicon.
- 🦅 swama — High-performance MLX inference engine for macOS with native Swift.
- 🐍 mlx-llm — LLM applications and tools running on Apple Silicon in real-time.
- 🐍 fastmlx — High-performance production-ready API to host MLX models.
- 🦅 PicoMLXServer — The easiest way to run MLX-based LLMs locally.
- 🐍 mlx-openai-server — OpenAI-compatible endpoints for MLX models.
- 🦅 maclocal-api — macOS server exposing Apple Foundation and MLX Models through unified OpenAI-compatible API.
- 🐍 mlx_parallm — Fast parallel LLM inference for MLX.
- 🐍 mlx-gui — MLX inference server with web UI for Apple Silicon.
- 🐍 mlxserver — Simple inference server for the MLX library.
- 🐍 Toolio — GenAI & agent toolkit for Apple Silicon, JSON schema-steered structured output and tool-calling.
- 🐍 mlx_sharding — Distributed inference for MLX LLMs across multiple devices.
- 🐍 llamactl — Unified management and routing for llama.cpp, MLX and vLLM models.
- 🐍 Rapid-MLX — Fast local AI engine for Apple Silicon. 4.2x faster than Ollama, tool calling, prompt caching. OpenAI-compatible.
- 🐍 TransformerLab — Open source research environment for training, evaluating, and scaling models.
- 🐍 autoresearch-mlx — MLX port of Karpathy's autoresearch — autonomous AI research loops on Mac.
- 🐍 mlx-tune — Fine-tune LLMs on Mac: SFT, DPO, GRPO, Vision. Unsloth-compatible API.
- 🐍 mlx-lm-lora — Train Large Language Models on MLX with LoRA.
- 🐍 SiLLM — Simplifies training and running LLMs on Apple Silicon via MLX.
- 🐍 MLX-GRPO — Pure MLX-based GRPO training pipeline on Apple Silicon.
- 🐍 mamba.py — Simple and efficient Mamba (state-space model) in pure PyTorch and MLX.
- 🐍 mlx-snn — Spiking Neural Network library built natively on MLX.
- 🐍 mlx-gpt2 — GPT-2 from scratch in MLX. Educational.
- 🐍 rlx — Reinforcement learning framework based on MLX.
- 🐍 Vodalus-Expert-LLM-Forge — Dataset crafting with RAG + fine-tuning using MLX and Unsloth.
- 🐍 unsloth-buddy — Zero-friction LLM fine-tuning for agents; MLX on Apple Silicon.
- 🦀 pmetal — High performance LLM fine-tuning framework for Apple Silicon, written in Rust.
- 🟨 voicebox — The open-source voice synthesis studio.
- 🐍 mlx-audio — TTS, STT and STS library built on MLX.
- 🐍 lightning-whisper-mlx — Extremely fast Whisper optimized for Apple Silicon.
- 🐍 parakeet-mlx — Nvidia's Parakeet models for Apple Silicon.
- 🐍 TheWhisper — Optimized Whisper for streaming and on-device use.
- 🐍 f5-tts-mlx — F5-TTS in MLX.
- 🐍 Lightning-SimulWhisper — MLX/CoreML streaming Whisper, ~15x faster.
- 🦅 speech-swift — AI speech toolkit — ASR, TTS, speech-to-speech, VAD, diarization.
- 🦅 mlx-audio-swift — Modular Swift SDK for audio processing with MLX.
- 🐍 qwen3-tts-apple-silicon — Qwen3-TTS on Mac. Voice cloning, 100% offline.
- 🐍 csm-mlx — Conversation Speech Model for Apple Silicon.
- 🐍 MLX-Auto-Subtitled-Video-Generator — Generate accurate video transcripts using MLX.
- 🐍 wtm — Blazing fast Whisper Turbo for speech-to-text on Mac.
- 🦅 kokoro-ios — Kokoro TTS for iOS and macOS via MLX.
- 🐍 whisply — Fast CLI/GUI for batch transcription and translation.
- 🐍 nanospeech — Simple, hackable TTS in PyTorch and MLX.
- 🐍 mflux — MLX native Flux and Stable Diffusion image generation.
- 🐍 mlx-video — Image-Video-Audio generation models on Mac.
- 🦅 flux.swift — Swift implementation of Flux.1 using mlx-swift.
- 🐍 mlxstudio — MLX Studio — Image Gen/Edit + Chat/Code all in one.
- 🐍 MFLUX-WEBUI — Web UI for MFLUX image generation using MLX and FLUX models.
- 🦅 flux-generator — Local image and music generation for Apple Silicon.
- 🐍 mlx-vlm — Vision Language Models on Mac using MLX.
- 🐍 ml-aim — Apple's AIMv1 and AIMv2 vision models.
- 🐍 pvm — Phi-3.5 Vision and Language Models for Mac.
- 🐍 photo-similarity-search — CLIP-based photo similarity for Apple Silicon.
- 🐍 mlx-embeddings — Vision and Language Embedding models locally on Mac.
- 🦅 VecturaKit — Swift vector database for on-device RAG.
- 🐍 jina-grep-cli — Semantic grep powered by Jina embeddings (MLX).
- 🐍 sisi — Semantic image search CLI using MLX embeddings.
- 🐍 mlx-retrieval — Train embedding and reranker models on Apple Silicon.
- 🐍 mlx-rag — Simple RAG application running locally on Apple Silicon.
- 🐍 mlx_clip — CLIP on Apple Silicon using MLX.
Native macOS/iOS apps and packages built on MLX. See also: mlx-swift in Core, swama in Inference, mlx-audio-swift in Audio. Getting started? Read On-device ML research with MLX and Swift.
- 🦅 osaurus — Native macOS AI agent — any model, persistent memory, autonomous execution.
- 🦅 Klee — Native macOS AI chat, 100% local inference.
- 🦅 ChatMLX — High-performance chat app for macOS.
- 🦅 Fabric — Creative Coding / 3D / Image Processing, inspired by Quartz Composer.
- 🦅 mlx-swift-chat — Multi-platform SwiftUI frontend for local LLMs.
- 🦅 fullmoon-ios — Chat with local LLMs on iPhone, iPad, Mac.
- 🦅 LocalLLMClient — Swift package for local LLMs on iOS, macOS, Linux.
- 🦅 MLX-Outil — Tool calling using MLX Swift across iOS, macOS, and visionOS.
- 🦅 mlx-swift-audio — Swift tools for TTS and STT powered by MLX.
- 🦅 f5-tts-swift — F5-TTS implementation in Swift using MLX.
- 🦅 Metal-Puzzles — Solve puzzles. Learn Metal GPU programming.
- 🐍 mlx-benchmark — Benchmark MLX ops on all Apple Silicon chips (GPU, CPU) + MPS and CUDA.
- 🐍 mlx-bitnet — 1.58 Bit LLM on Apple Silicon.
- 🐍 chat-with-mlx — All-in-one LLMs Chat UI for Apple Silicon.
- 🐍 cross-market-state-fusion — RL agent fusing Binance futures into Polymarket. On-device training with MLX.
- 🦅 NotebookMLX — Open source NotebookLM.
- 🟨 nodetool — Visual builder for AI Workflows and Agents.
- 🐍 Vim-LM — AI Copilot for Vim/NeoVim.
- 🐍 mlx-ui — Simple web UI for MLX using Streamlit.
- 🟨 Silicon-Studio — Fine-tune and run LLMs locally on M-series Mac.
- 🐍 llm-mlx — MLX model support plugin for Simon Willison's LLM tool.
- 🐍 mlx-chat-app — macOS chat app connecting local docs to MLX-powered LLMs.
- 🐍 jarvis-mlx — All-in-one offline productivity assistant using MLX.
- 🦅 AirPosture — AirPods as AI posture coach on iOS, powered by MLX.
- 🐍 nanoGPT_mlx — Karpathy's nanoGPT on Apple MLX.
- 🐍 m-courtyard — Local AI model fine-tuning assistant, zero-code.
- 🦀 llmfit — One command to find what LLM runs on your hardware.
- 🐍 einops — Flexible tensor operations (supports MLX).
- 🐍 outlinesmlx — Guided generation on Apple Silicon using Outlines + MLX.
- 🐍 olla — Lightweight proxy/load balancer for LLM infra (llama.cpp, MLX, vLLM).
- 🐍 anubis-oss — Local LLM testing and benchmarking for Apple Silicon.
Where to find models:
- mlx-community on HuggingFace — 2000+ MLX-optimized models ready to use
- lmstudio-community on HuggingFace — LM Studio's curated MLX models
- Using MLX at Hugging Face — Official HuggingFace MLX docs
Which model for my Mac?
| RAM | Recommended models | Notes |
|---|---|---|
| 8GB | Qwen3.5-4B-4bit, Phi-4-mini-4bit | Small but capable |
| 16GB | Qwen3.5-9B-4bit, Gemma-3-12B-4bit | Great balance |
| 32GB | Qwen3.5-27B-4bit, Devstral-24B-4bit | Strong coding + reasoning |
| 64GB+ | Qwen3.5-122B-A10B-4bit, Llama-4-Scout | Near-frontier quality |
- On-device ML research with MLX and Swift — Official Swift blog (2024)
- Deep Dive into AI with MLX and PyTorch — Comprehensive educational resource
- Deploying LLMs locally with MLX — Toward Data Science
- MLX Community Projects — Official MLX discussion thread for community projects
Built something with MLX? We want to list it!
Easiest way: Open an issue with your project name, URL, and a one-line description. We'll add it for you.
PR way: Add your project to the right category using this format:
- 🐍 [name](url) — One-line description. 
Add a badge to your README (optional):
[](https://github.com/raullenchai/awesome-mlx)Guidelines:
- Language emoji: 🐍 Python · 🦅 Swift · 🟨 JS/TS · 🦀 Rust · 🐹 Go
- Sort entries by stars (descending) within each section
- All MLX projects welcome — no minimum star count