A.I. designed for you, only for you.
Run powerful language models locally on iPhone and Mac. Keep your conversations private. Your devices, your models, your data.
See Arbiter on iPhone
Why cloud AI falls short
Your data leaves your device
Every prompt you send to a cloud AI is processed on someone else's servers, logged, stored, and potentially used to train future models.
Requires internet to function
Cloud AI stops the moment your connection drops. On a flight, in a dead zone, or on a slow network, you are left without assistance.
Expensive and metered
Premium AI is locked behind monthly subscriptions and per-query fees. Your usage is tracked, throttled, and billed at every turn.
Key Features
On-Device AI
Run open-source models directly on your iPhone or Mac using both GGUF and MLX frameworks on either platform. Chat offline, analyze images with vision models, and keep everything on your hardware. No API calls, no server round-trips.
Live Web Search
Combine on-device inference with live web results. Arbiter pulls up-to-date information from the web and reasons over it locally, so your searches stay private while your answers stay current.
44 Model Catalog
Browse, download, and switch between 44 models from families like Gemma, Llama, DeepSeek, Qwen, Mistral, and Phi. Both GGUF and MLX formats run on iPhone and Mac, with 9 vision and 10 reasoning models available.
Files & Smart Context
Drop in PDFs or text files and ask questions about them. Arbiter manages context windows intelligently across models with different limits, summarizing older messages to keep conversations flowing.
Apple Foundation Models
On devices with Apple Intelligence, use Apple’s on-device Foundation Models with zero downloads. Available on iOS 26 and macOS 26, integrated directly into Arbiter’s chat experience.
Local Network Models
Connect your iPhone to a Mac, PC, or any device running LM Studio, Ollama, or Arbiter’s own server over local Wi-Fi. Run larger models on powerful hardware and chat from your phone.
Mac as Model Server
Turn your Mac into a private AI server. Serve installed MLX models through an OpenAI-compatible API and share them with your iPhone or any compatible client on your network.
Siri & System Integration
Trigger Arbiter hands-free via Siri Shortcuts using App Intents. Ask questions by voice and get spoken or text responses, all private and built into your Apple workflow.
Roles & Personalization
Choose from 10 built-in assistant roles like Coding Helper, Translator, Meal Planner, and Study Buddy. Fine-tune warmth, enthusiasm, emoji usage, and response detail to match your style.
Arbiter vs. Cloud AI
Our Mission
Arbiter is redefining the future of AI chat across iPhone and Mac. We believe in empowering individuals with tools that offer choice, control, and clarity. By running AI locally and giving users ownership of their data and hardware, Arbiter puts privacy and power back into your hands.
FAQ
Is my data private?⌄
Yes. Everything is processed and stored on your device. Your chats and files never leave your phone or Mac unless you explicitly choose to share them. This applies across iPhone and Mac — whether running models locally or serving them on your network, data never touches external servers.
Do I need an internet connection to use the app?⌄
No. Once a model is installed, the app works entirely offline. You can use it on airplanes, in remote areas, or without Wi-Fi, while still getting fast and responsive answers.
How does offline AI work?⌄
Arbiter uses optimized open-source models that run locally on your device’s Apple Silicon processor. Both GGUF and MLX model formats are supported on iPhone and Mac. Models are pre-downloaded and don’t require server access, so your requests are processed instantly without sending anything to the cloud.
What devices are supported?⌄
Arbiter runs on iPhones and iPads with iOS 16 or later, with best performance on devices with A14 Bionic or newer and at least 6 GB of RAM. Arbiter for macOS is available on Apple Silicon Macs (M1 or later) and can run larger MLX models.
What AI models are available?⌄
Arbiter offers 44 models across families including Gemma, Llama, DeepSeek, Qwen, Mistral, Phi, and Granite. The catalog includes 24 GGUF and 20 MLX models, all usable on both iPhone and Mac, plus 9 vision-capable models and 10 reasoning models. On supported devices, Apple’s Foundation Model is also available.
What is Apple Foundation Model support?⌄
Starting with iOS 26 and macOS 26, Arbiter integrates Apple’s on-device Foundation Models through Apple Intelligence. These models run natively on your device at no additional cost, providing another option alongside open-source models with zero downloads required.
Can I use my Mac as a server for my iPhone?⌄
Yes. The macOS app includes a Serve Model feature that creates an OpenAI-compatible local API server. Your iPhone can connect to your Mac over Wi-Fi using automatic Bonjour discovery, letting you run larger models on your Mac while chatting from your phone.
Can I connect to other local AI servers?⌄
Yes. Beyond Arbiter’s own Mac server, you can connect your iPhone to any local machine running LM Studio, Ollama, or any OpenAI-compatible API server on your local network. Configure the host and port and you’re connected.