Dunesies Personal

A native macOS and iOS app that runs large language models directly on your own machine. Pick a model, load it once, and every question, document and line of code you share is answered by your own hardware, with no account, no subscription and nothing leaving your computer. Which means you can finally ask about the files you would never upload anywhere: the contract, the medical record, the salary sheet, the code that isn't public yet.

Hi

Hello, I'm Dunesies. I run locally on your device.

Nice, so nothing I say leaves my Mac?

Exactly. No cloud, no accounts. It all stays right here.

How much does this cost?

Dunesies itself is free. You just need your own Mac to run it on.

Can I switch models anytime?

Anytime. Download any model you like and switch in seconds.

Send a message

Overview

Everything you need, on your machine.

Fully local inference

Models run on your Mac through llama.cpp. Nothing is uploaded, nothing is logged remotely, and no connection is needed once a model is on disk.

Automatic setup

The app fetches a prebuilt llama.cpp engine in about thirty seconds. No Terminal, no Xcode tools, no compiler prompts.

Curated model library

Nineteen ready to run models, one click each, plus any GGUF model from a Hugging Face URL or a file already on your disk.

Hardware aware guidance

The app reads your real RAM and chip, then labels every model before you spend time downloading it.

Isolated chat sessions

Each conversation runs in its own process, so context never bleeds between chats. History is restored on relaunch.

File attachments

Attach PDFs, source files, CSVs, logs and markdown, or images when running a vision model, and ask about them directly.

Real time streaming

Answers stream token by token as they are generated, so you watch the response form instead of waiting on a block of text.

Searchable history

A sidebar keeps every session to hand, with search across titles and one click switching between conversations.

Requirements

What you need to run it.

macOS

macOS 14.0 or laterUniversal build for Apple Silicon and Intel

Memory

8 GB minimum, 16 GB recommendedmacOS reserves 2.5 GB on Apple Silicon and 3.5 GB on Intel when rating a model

Disk space

About 2 GB, plus each modelModels range from 1.6 GB to 40 GB

Internet

Setup and downloads onlyInference itself is fully offline

Cost

FreeNo account, no subscription, no per token billing

iOS

iOS 17.0 or lateriPhone and iPad, with smaller models sized to fit

Getting started

From download to first answer.

Eight steps, and only the first two happen more than once.

01

Install the app

Open the .dmg and drag Dunesies into your Applications folder.

02

Set up the engine

This is automatic. The app downloads a prebuilt llama.cpp engine on first launch, around thirty seconds, once.

03

Pick a model that fits your machine

The Models tab rates every model against your real hardware, so start with anything marked Runs Great. 7B or 8B is the sweet spot on 16 GB machines.

04

Download it once

Click download and wait for the file to land, paste a Hugging Face GGUF link, or choose a .gguf file already on your disk.

05

Start a chat

Switch to the Chat tab, choose your model from the pill at the top, and type. Answers stream back as they generate. The plus button starts a new session.

06

Bring in your own files

Use the paperclip to attach a PDF, source file, CSV or log and ask about its contents. Vision models accept images too.

07

Work offline

Once a model is downloaded you can disconnect entirely. On a plane, on a train, or on a locked down network, it works exactly the same.

08

Switch models whenever you like

Keep several models installed and change between them from the picker at the top of the chat. Each session remembers what it was started with.

Comparison

Why it beats renting intelligence.

Cloud assistants charge monthly, send your work to someone else's servers, and stop the moment the connection or the subscription does. It beats running llama.cpp yourself too: no repository to clone, no compiler to install, no source to build, just the same proven engine in a real desktop app.

	Dunesies Personal	Cloud AI subscriptions
Where your data goes	Never leaves your Mac	Uploaded to third party servers
Ongoing cost	Free, no subscription	Monthly fee, often per seat
Usage limits	None, bound only by your hardware	Message caps and rate limits
Works offline	Yes, after the model downloads	No, requires a connection
Account required	No	Yes
Choice of model	Any GGUF model you want	Whatever the vendor offers
Model changes under you	Never, your copy stays fixed	Updated or retired without notice
Training on your content	Impossible by design	Depends on vendor policy
Setup effort	One click, about 30 seconds	Sign up and billing

Models

Any GGUF model you want.

Nineteen models ship in the library, each one click and quantized to Q4_K_M. You can also bring your own file, and the same formats apply.

Model	Parameters	Download	Best for
Gemma 2 2B Instruct	2B	1.6 GB	Quick responses, low resources
Qwen2.5 3B Instruct	3B	1.9 GB	General chat, quick responses
Llama 3.2 3B Instruct	3B	2.0 GB	General purpose, fast
Phi-3 Mini 3.8B	3.8B	2.2 GB	Reasoning, coding, chat
Phi-3.5 Mini 3.8B	3.8B	2.3 GB	Reasoning, efficient performance
Qwen2.5 7B Instruct	7B	4.4 GB	General purpose, versatile
Mistral 7B Instruct v0.3	7B	4.4 GB	Chat, reasoning, coding
Qwen2.5 Coder 7B	7B	4.7 GB	Coding, lightweight
Llama 3.1 8B Instruct	8B	4.9 GB	General purpose, versatile
Gemma 2 9B Instruct	9B	5.4 GB	Reasoning, math, science
CodeLlama 13B Instruct	13B	7.9 GB	Code generation, debugging
Qwen2.5 14B Instruct	14B	8.5 GB	Advanced chat, analysis, coding
Gemma 2 27B Instruct	27B	16 GB	Advanced tasks, high quality
Qwen2.5 32B Instruct	32B	19 GB	Professional tasks, complex reasoning
Qwen2.5 Coder 32B	32B	19 GB	Advanced coding, debugging, refactoring
DeepSeek Coder 33B	33B	19 GB	Professional coding, architecture design
Mixtral 8x7B Instruct	47B MoE	26 GB	Efficient large model, strong general reasoning
Nous Hermes 2 Mixtral 8x7B	47B MoE	26 GB	Enhanced chat, instruction following
Llama 3.3 70B Instruct	70B	40 GB	Professional use, complex reasoning

Technical

Under the hood.

Inference engine

llama.cppDownloaded automatically on first launch

Acceleration

MetalMetal on Apple Silicon with offload tuned to your RAM, with a CPU fallback

Context window

8,192 tokensClamped to the model's own maximum when smaller

Model format

GGUFCurated library, Hugging Face URLs, or local files

Session isolation

One process per chatNo context bleeding between conversations

Take back control.

No account, no subscription, free to keep. Your models, your machine.

← All applications