OnDevice LLM - Offline AI Chat

About this app

<p>OnDevice LLM runs powerful open-source AI models 100% on your iPhone — no internet, no cloud, no account.</p>
<p>Your conversations never leave the device. Every word you type, every reply you get, stays on your phone. There is no server to log it, no company to mine it, no account that can be banned.</p>
<p>NEW IN 1.3 — TERNARY BONSAI, HOME SCREEN, DAILY BRIEF<br />
- Ternary Bonsai arrives: 1.7B / 4B / 8B (MLX) — 8B-class quality in just 1.75 GB, fully offline.<br />
- Brand-new Home screen with Quick Actions: ask a document, ask a photo, brief my day, or start a chat.<br />
- Brief my day: combine your Calendar + Reminders with the app's private memory into a morning briefing — fully on-device.<br />
- Better DeepSeek R1 output (fixed chat template).<br />
- Model info cards on every model — check size, fit, and format before loading.</p>
<p>WHAT'S INSIDE<br />
- A curated catalog of top open models:<br />
- llama.cpp / GGUF: Gemma 4 / Gemma 2, Llama 3.2, Qwen 2.5 / Qwen 3, Phi-4 Mini, SmolLM2, DeepSeek R1.<br />
- MLX (Apple Silicon native): Ternary Bonsai 1.7B / 4B / 8B.<br />
- Quick Start chips: tap a persona (Translator, Code Helper, Brainstorm, Email Writer, ELI5...) or a demo (Rock Paper Scissors, Click Counter, Snake, Calculator, Tic-Tac-Toe, Drawing Pad, Bouncing Ball, SVG illustration) to send the prompt with one tap.<br />
- Run mini-apps in chat: ask for a game, calculator, drawing pad, or SVG and tap "Run" to use the result right inside the chat — fully offline.<br />
- Markdown chat with code blocks, syntax highlighting, copy + share.<br />
- Attach PDFs, images, or text documents — the AI reads them locally.<br />
- Voice input via on-device speech recognition.<br />
- Persistent chat history with search.<br />
- Personal memory that learns useful facts about you (and stays on your phone).<br />
- Share Extension: send text from any app straight to your AI.</p>
<p>PRIVACY BY DESIGN<br />
- 100% offline inference. Airplane mode works. Cellular off works.<br />
- No account. No email. No tracking SDKs. No analytics that report your prompts.<br />
- Your chats, memories, and downloaded models live in your iPhone's app sandbox only.</p>
<p>WHO IT'S FOR<br />
- Privacy-conscious professionals (legal, medical, journalism, therapy notes).<br />
- Frequent travelers, pilots, mariners — anyone working without reliable internet.<br />
- Users in countries where cloud AI services are blocked.<br />
- Tinkerers and developers who want to load and compare open-source models on a phone.</p>
<p>PRO UPGRADE<br />
- Free plan: 1 model loaded, 3 saved chats.<br />
- Pro: unlimited models, unlimited chats, custom system prompts, custom HuggingFace model URLs.<br />
- One-time Lifetime purchase — buy once, own forever.</p>
<p>Subscriptions auto-renew unless canceled at least 24 hours before the period ends. Manage or cancel any time in Settings - Apple ID - Subscriptions.</p>
<p>Terms of Use (EULA): https://www.apple.com/legal/internet-services/itunes/dev/stdeula/</p>