Edge Mind Ai is a privacy-first AI assistant that runs entirely on your iPhone and iPad. Your conversations, prompts, and photos never leave your device — all AI inference happens locally on Apple Silicon. There is no cloud server, no account required, and no data collection.
RUN 20+ MODELS, LOCALLY
Choose from a curated catalog of 21 open models from Apple Intelligence, Google Gemma, IBM Granite, Meta Llama, Microsoft Phi, DeepSeek, Mistral, Qwen, SmolLM, and Liquid AI. Switch between them mid-conversation. Download only the ones you want; delete them anytime to reclaim space.
REAL CHAT, NOT A WRAPPER
• Token-by-token streaming responses
• Full Markdown rendering with code blocks and syntax highlighting
• Multiple conversations — create, rename, and resume sessions from History
• Attach a photo from your camera or library and ask a vision model about it
• Voice dictation — speak your prompt; responses read back via on-device speech
PRIVACY IS THE DEFAULT
• Fully offline out of the box. Nothing is uploaded unless you turn on Live Search.
• No telemetry, no analytics, no cloud sync, no login required.
• A guest profile gets you chatting in one tap.
• Optional Face ID / Touch ID / local credentials to protect your space.
• Your HuggingFace token (for gated model downloads) is stored in the iOS Keychain.
OPTIMIZED FOR YOUR DEVICE
Edge Mind Ai detects your chip and tunes itself automatically — context windows of 2K–8K tokens and flash-attention acceleration on supported devices, so large models run within your iPhone's memory without crashing.
OPTIONAL LIVE WEB SEARCH
Turn on web grounding per prompt when you need fresh information. Bring your own key for Tavily, Brave Search, Serper (Google), or a self-hosted gateway. Search results come back with inline source citations. It's off by default — your AI stays local unless you ask it to look something up.
CURATED, NOT THE WILD WEST
The model catalog is hand-picked and capability-audited for this app's runtime. Thinking models stream a reasoning lane; vision models accept images; tool-calling models can trigger web search. Badges reflect what the app actually supports, not just marketing claims.
WHO IT'S FOR
• Privacy-conscious users who want private AI assistance without sending data to a server
• Developers and researchers experimenting with local LLMs on iOS
• Anyone who needs an AI assistant that works on a plane, underground, or offline
• People who want to own their AI — models live on your device, under your control
REQUIRES iPhone or iPad with iOS 17 or later and an A14 chip or newer for best results. MLX models run on physical Apple Silicon only (not the simulator). Apple Intelligence model requires a compatible device.
Edge Mind Ai is an independent project. It is not affiliated with or endorsed by Apple, Google, Meta, Microsoft, IBM, Alibaba, Mistral, Hugging Face, or Liquid AI. Model weights are the property of their respective creators and are loaded under their original licenses.
평가 및 리뷰
이 앱은 개요를 표시할 만큼 충분한 리뷰 또는 평가를 받지 않았습니다.
• A fresh new look. Chat, models, documents and settings have been redesigned with a calm graphite design, glass controls and a new serif typeface, in both dark and light mode.
• New: "This iPhone" check. See which models run well on your iPhone, which are a tight fit, and which are too large, before you download anything.
• Every download now shows how well the model fits your iPhone, how much space it needs, and what it can do, and suggests a model that fits better when one is too large.
• Switch models from the chat header, start a chat with one-tap suggestions, and have answers read aloud or share them.
• Answers look better: code blocks with a Copy button, cleaner tables and quotes.
• Reasoning shows as one tidy line you can expand, and replies are no longer mistaken for reasoning on small Qwen models.
• A redesigned Memory page and a new welcome screen.
• New: Reasoning setting (Settings → AI Configuration). Choose how long reasoning models such as Qwen and DeepSeek R1 think before they answer: Off, Short, Medium or Long. Short is the new default, so these models reply much faster.
• Model downloads now keep going in the background for every model type, including MLX. Switch apps or lock your phone and the download continues. Progress shows real sizes, and Cancel leaves nothing behind.
• Fixed: Gemma models could stop responding after using a tool or tapping Stop until the app was restarted.
• Better answers from your documents: follow-up questions about an attached file keep working, library search only uses relevant passages, and long documents keep their full text after you restart the app.
• Models now remember what you told them earlier in long conversations, including Apple Intelligence and Gemma.
• Saved memories are only used when they are relevant to your question.
• Calculations and questions about the time, battery or your device are answered reliably on every model.
• Qwen 3.5 models now reason properly before answering.
• Many other fixes and stability improvements.
버전 0.3.5
Vinoth Rajalingam 개발자가 아래 설명된 데이터 처리 방식이 앱의 개인정보 처리방침에 포함되어 있을 수 있다고 표시했습니다. 자세한 내용은 개발자의 개인정보 처리방침 을 참조하십시오.
데이터가 수집되지 않음
개발자가 이 앱에서 데이터를 수집하지 않습니다.
개인정보 처리방침은 사용하는 기능이나 사용자의 나이 등에 따라 달라질 수 있습니다. 더 알아보기
