Minimalist AI Pro brings chat, image generation, transcription and speech to your iPhone, iPad and Apple silicon Mac — running on your device. Download a recommended model or import your own, then keep working offline.
Private by design
Your chats, prompts, images, transcripts and audio stay on your device. They are not sent to a cloud AI service. No account, no ads, no tracking.
Three engines, your models
GGUF models (llama.cpp), MLX models for Apple silicon and MiniCPM-V packages all run inside the app. Recommended downloads show their size and memory guidance, and several downloads can run, pause and resume side by side. You can also import compatible GGUF files, MLX folders and MiniCPM packages. Supports the latest MiniCPM-V-4.7-35B-A3B and Bonsai 2 (27B).
See, think and calculate
Ask about photos with vision models, turn on thinking for models that support it, and let optional JavaScript and Python tools do exact calculations while you inspect the code and the result. Python uses a lightweight built-in interpreter, not a full CPython environment.
Images from your words
Generate 512×512 images with Stable Diffusion 1.4 or 1.5 (Apple Core ML) or Bonsai Image 4B, or import Apple Core ML Stable Diffusion 1.x/2.x models. Stable Diffusion understands English prompts best. Every image passes a safety checker before it is shown or saved. Image generation needs a 12 GB-class iPhone or iPad, or a Mac with 16 GB of memory or more.
Transcribe and speak
Turn audio and video files into text on device, read text aloud, or create a WAV file to listen to and share. Transcription needs iOS, iPadOS or macOS 26 or later.
Make it yours
Chat, images and audio each keep their own drafts, background, transparency and bubble colors. Give the assistant a name, tell it yours, and add your own system prompt. Formulas, Markdown, tables and code are clearly rendered.
One purchase
A one-time purchase with no subscription and no in-app purchases. Models are not included; downloading them uses the network, and you decide whether cellular data may be used.
Answer quality, image results and speed depend on the model and your device, including available memory. AI can make mistakes. Check important information.
Show more
