AI on Device
AI on Device - Local Assistant is an app that offers advanced AI capabilities with an emphasis on privacy and ease of use. Leveraging the p...
Last verified:
What is AI on Device?
AI on Device - Local Assistant brings powerful large language models (LLMs) directly to your Apple devices, running entirely offline to ensure complete privacy and no internet dependency. It supports popular models like Meta Llama, Google Gemma, Microsoft Phi and Wizard, Mistral, Qwen, and YI, allowing seamless multi-round conversations, document analysis, and web browsing capabilities. Users can customize system prompts, model temperatures, and interact with PDFs, photos, and websites for quick insights, making it ideal for secure environments, remote locations, or anyone prioritizing data privacy.
Key features include advanced document interaction for extracting information and summaries, performance enhancements for smoother responses, Siri and Shortcuts integration, and tools for creative writing, academic research, coding assistance, language tutoring, and personal productivity. The app offers an intuitive interface with stop generation buttons, feedback forms, and background downloads for models. It's designed for developers, researchers, professionals, students, and curious users seeking flexible, local AI without data collection.
Compatible with iOS 16+, macOS 14+, iPadOS 16+, and visionOS 1.1+ on devices with A12 Bionic or later (iPhone 14+ recommended for best performance), the app emphasizes no user data collection, making it a truly private AI companion. Stay productive offline with essay generation, data analysis, wellness support, and more, while upcoming updates promise expanded capabilities.
AI on Device pricing
Pricing model: Free
Free to download and use with core features; freemium model with optional one-time paid upgrades from $3.99 for advanced capacities, performance enhancements, or additional models.
AI on Device pros
- Runs entirely offline on device
- Supports Meta Llama models locally
- Google Gemma integration for compact use
- Microsoft Phi and Wizard compatibility
- Mistral models for versatile tasks
- Qwen and YI model support
- Complete privacy with no data collection
- Free core app download
- Customizable system prompts
- Adjustable model temperature settings
- PDF and photo document analysis
- Web browsing expands model knowledge
- Siri and Shortcuts integration
- Multi-round conversation support
- Performance optimizations for smooth UI
- Background model downloads
- Stop generation button for control
- No internet needed for core features
AI on Device cons
- Requires A12 Bionic chip minimum
- iPhone 14+ recommended for best RAM
- Large model downloads (2-20GB+)
- Slower inference on lower RAM devices
- No support for older iOS/macOS
- Limited to Apple Silicon devices
- Models need manual download first
- No cloud AI hybrid by default
- VisionOS requires 1.1+ version
Frequently asked questions about AI on Device
What devices are compatible with AI on Device?
Compatible with iPhone (iOS 16.0+ on A12 Bionic or later, iPhone 14+ recommended), iPad (iPadOS 16.0+ on A12+), Mac (macOS 14.0+ Apple Silicon M1+), and Apple Vision Pro (visionOS 1.1+). Devices with 6GB+ RAM recommended for optimal model performance.
Does it require an internet connection?
No, it runs completely offline after model download. Core AI chat, document analysis, and conversations process locally without internet, ensuring privacy; web browsing feature requires connection to expand knowledge.
What AI models does it support?
Supports popular open-source LLMs like Meta Llama, Google Gemma, Microsoft Wizard and Phi, Mistral, Qwen, YI, and more. App recommends models based on device RAM, from compact 4B to large 32B parameter models.
Is user data collected or shared?
No data is collected from users. All processing happens on-device, with no chat history, personal info, or usage data shared with any service, prioritizing complete privacy.
How do I download and use a model?
Open the model picker in AI Chat, browse recommended models for your hardware, tap Download (sizes 2GB-20GB+), wait for completion, then start chatting. No setup needed post-download.
Can it analyze documents and PDFs?
Yes, enhanced document interaction allows reading PDFs and photos directly for quick information extraction, detailed analysis, summaries, key points, and data insights with AI-driven visualizations.
What customization options are available?
Advanced system settings let you customize model behaviors, system prompts for tailored responses, temperature for creativity/randomness, and preferences like voice/speech in settings.
Does it integrate with Siri or Shortcuts?
Yes, supports Siri and Apple Shortcuts for running LLMs directly, enabling automation, app management, and seamless workflow integration on iOS and macOS.
What tasks can it help with?
Creative/academic writing, question answering, language tutoring/translations, document analysis, coding/debugging, productivity tools like to-do lists, event planning, and wellness advice.
What are the latest updates?
Version 0.7.1 adds background downloads, performance/UI updates, more models, and context embedding adjustments to match token capacity for better interactions.