Know your Mac
aituner reads the chip, the GPU and the unified memory, and benchmarks what your Mac can really do for AI.
self-hosted ai on your mac
Your Mac. Your private AI. Tuned.
aituner measures your Mac, finds the most capable model that fits its memory and runs it in one click, with a private chat. No Terminal, no cloud: your data stays on your Mac.
Or the notarized edition with Homebrew:
brew install --cask aituner-app/tap/aitunerFree, for Apple Silicon Macs. New to local AI? Self-hosted AI on a Mac.
how it works
From your Mac to your own AI in one click
Press Get started. aituner does the rest, step by step, and shows you each one.
01 · measure
02 · fit
The most capable model for it
Choose AI Performance (the most capable model) or AI Memory (the longest context window). A ring shows how the memory splits between the model and its context.
03 · run
Chat with it, privately
aituner downloads the MLX model, runs it inside the app and opens Local AI, a chat with it on your Mac.
private by design
Powerful AI that never leaves your Mac
reviewed
Reviewed by Apple
Each version is checked against Apple's App Review Guidelines before it is published.
sandboxed
In Apple's App Sandbox
It can reach only what macOS allows it to. Everything it needs to run models is inside the app.
one account
One private account
Sign in with Apple once. Your account knows who you are and which devices are yours, nothing more.
private
Stays on your Mac
Models run on your Mac and your chats are kept there. No tracking, no advertising.
your private aituner account
Your account holds the keys. Your Mac holds everything else.
Sign in with Apple once. At account.aituner.app you see every Mac you own, live, from anywhere: what is running, the model it serves, GPU, CPU and memory, and the iPhones and browsers connected to it. Nothing your Mac holds is ever stored in the cloud.
your account · cloud
Who you are, which devices are yours
- Apple's id for you, your email or Apple's relay address
- Your Macs, iPhones and browsers, and when each was last seen
- Your support tickets' numbers
your mac · the vault
Everything else, only on your Mac
- Your chats and the files your model reads
- Your models, benchmarks and results
- Live GPU, CPU and memory, and what is running
devices
All your Macs and iPhones
Every device on your account, when it was last seen, and Disconnect or Remove for any of them.
live · plus
Your Mac, from anywhere
Its sessions and model live, and a chat with it from any browser, end to end encrypted. Free during the beta.
feedback
Built on your feedback
Tell us what to fix or build next from your account, and follow it to the release. Tickets are public issues on GitHub, with no name or email.
models
Every model, drawn against your Mac's memory
Each MLX model as a bar against your Mac's memory at the context you pick: weights, KV cache and the room left. Fits, tight or won't fit; run any in one click, or add one from Hugging Face. How memory decides what fits.
local ai
A private chat with the models on your Mac
Local AI talks to the model aituner runs, or one loaded in Ollama or LM Studio; the edition installed with Homebrew also reaches MLX servers and llama.cpp. Nothing goes to a cloud service.
sessions
Kept on your Mac
Several conversations, kept on this Mac only.
attach
Your documents
Attach text files and PDFs to any model and ask about them, on your Mac.
reasoning
Thinking apart
A thinking model's reasoning is shown apart from its answer.
live
Your Mac as it answers
GPU, CPU, memory and speed beside the chat.
menu bar
Always one click away
What runs (Ollama, LM Studio, MLX) and its model, a benchmark as it runs, GPU, CPU and memory live, and Stop all.
inside aituner
Every number, from your Mac
what it measures
Everything that decides how fast local AI runs
Real measurements on your Mac, never estimates: the median of several trials, saved with the thermal state and tool versions.
01 · discover
Hardware
Chip, performance and efficiency cores with their clocks, GPU cores, unified memory, the GPU memory limit, rated and measured bandwidth, thermal state, free disk.
02 · benchmark
Performance
Memory bandwidth, GPU compute (fp16 and fp32), prompt and generation speed with MLX on a fixed model, and Ollama when it is running. Live, with a log.
03 · compare
Results
Results shows how your Mac stands against the same chip and memory elsewhere.
FAQ
Questions
Can I run an LLM locally on my Mac without the Terminal?
Yes. In aituner from the Mac App Store, Get started benchmarks your Mac, then downloads the most capable MLX model for it and runs it inside the app: Apple's MLX is built in, so there is no Terminal and no Python. You sign in with Apple once, then chat with it in Local AI.
Which LLM fits my Mac?
It depends on its unified memory, the context window you want and what else is open. aituner's Models screen draws each MLX model against your Mac's memory at 16K, 32K or 128K tokens of context: its weights, its KV cache and the room left, with fits, tight or won't fit.
Does aituner work with Ollama and LM Studio?
Yes. Local AI chats with the model aituner runs and with any model loaded in Ollama or LM Studio; the edition installed with Homebrew also reaches MLX servers and llama.cpp. Active shows which of them are running and the models they have loaded. You do not need any of them: aituner runs MLX models itself.
Is aituner free?
Yes. aituner for Mac is free to download from the Mac App Store, with no license key. You sign in once with Apple, and AiTuner Plus is free during the beta.
Why do I sign in?
Your free aituner account ties your Macs, iPhones and browsers together: at account.aituner.app you see your Macs live, chat with their models from anywhere, manage your devices and follow your support tickets. Sign in with Apple handles it: no password of ours, and you can hide your email.
Can you see my chats?
No. Your account keeps who you are and which devices are yours. Your chats, models, results and live stats stay on your Mac; your own iPhone and browser see them end to end encrypted through our relay, which cannot read them and stores nothing.
Is aituner safe to install?
aituner for Mac comes from the Mac App Store: Apple reviews each version against its App Review Guidelines before it is published, it runs inside Apple's App Sandbox, so it can reach only what macOS allows it to, and it is installed and updated by the App Store. Your chats stay on your Mac, and there is no tracking or advertising.
What does private mean here?
The models run on your Mac. What you type to them, and what they answer, stays on your Mac: your conversations are kept there only, and nothing is sent to an AI company.
Which Macs does aituner support?
Apple Silicon Macs (M1 and later) with macOS 14 or later. Local AI runs on the Apple Silicon GPU, so that is what aituner measures and uses.
Does a benchmark change my Mac?
No. The benchmarks only measure. aituner downloads the fixed test model it needs (about 2 GB) once, after showing you what it will fetch.