Software to Run
Local AI on Your Mac

Beautiful desktop app with a built-in model browser. Browse hundreds of models, download with one click, and start chatting instantly. Feels like using ChatGPT — but everything runs on your Mac. The 0.4.x releases added Bionic, a companion app that turns any local model into a file-editing, script-running agent. Since version 0.4.14 it supports multi-token-prediction (MTP) speculative decoding, which speeds up generation on models that ship MTP heads, such as the Qwen 3.6 family.

Open-source desktop app designed to feel like ChatGPT. Clean conversation threads, model management, and full offline capability. Great if you want a familiar chat experience without cloud dependency.

Nomic's free desktop chatbot focused on simplicity. Standout feature: point it at a folder of PDFs or notes and ask questions about your own documents — all processed locally.

Modern desktop app that connects to both local models and cloud APIs. Supports Ollama models, online providers, and lets you compare outputs side-by-side. Beautiful Mac-native design.

Open-source web app you run locally that connects to Ollama, OpenAI, or any API. Beautiful ChatGPT-like interface with conversation history, prompt templates, and markdown rendering. Easy one-click deploy.

Lightweight tool that downloads and runs models with a single command — since v0.32, typing ollama opens an interactive local coding agent. v0.40 (4 October 2026) runs every model architecture its MLX engine supports on MLX by default on Apple Silicon, and v0.35 added decision models — small models such as Nimble that return choices and probabilities instead of text, for triage and routing. The MLX engine uses the M5 Neural Accelerators for much faster prompt processing, accepts image input and supports DFlash speculative decoding. Runs as a background service — many GUI apps connect to it.

Open-source (Apache 2.0) menu-bar app and server built on Apple's MLX. It pages the KV cache to your SSD, so returning to a long conversation skips recomputing it — the project cites 90 seconds down to under 5 on a context shift — serves several models with continuous batching, and speaks both the OpenAI and Anthropic APIs, which is why agent tools point at it. Its Lightning MTP mode runs multi-token-prediction speculative decoding on models that ship MTP heads; MacStories used it for its M5 Ultra review.

Beautiful SwiftUI app built specifically for Mac and iPhone. Connects to Ollama running on your Mac and gives you a polished, Apple-native chat experience. Also works on iPad and iPhone.

Comprehensive platform combining local AI chat with document management, workspaces, and agents. Upload PDFs and web pages, then chat with them using your local models.

Optimized inference toolkit for deploying LLMs locally with quantization support. Achieves up to 1.8x faster inference than standard Ollama through efficient memory management and batched processing. Great for running multiple concurrent requests.

Apple's machine learning framework designed from the ground up for Apple Silicon. According to the LLMCheck index, MLX achieves the highest possible performance by fully leveraging unified memory, Metal GPU, and the Neural Engine.

The C/C++ inference engine powering most local AI tools under the hood. Gives you full control over quantization, context lengths, batch sizes, and Metal GPU offloading.

Feature-rich web UI that connects to Ollama or any OpenAI-compatible API. Supports multiple users, RAG document chat, web browsing, image generation, and plugin extensions.

Revolutionary tool that lets you pool multiple Apple Silicon Macs together over your local network to run larger models. Two M4 Max 64GB Macs become a single 128GB inference cluster. Perfect for running 70B+ models without a single high-RAM machine.