LM Mini vs HAIchat

Chat with LM Studio, Ollama, and oMLX from your iPhone — or run on-device models offline. Private local AI on hardware you control. No cloud chat subscription required. THREE WAYS TO USE IT 1. On-device — download a model and chat with no internet 2. Connected — LM Studio, Ollama, Jan AI, or oMLX over Wi‑Fi or USB 3. Remote — LM Mini Connect from anywhere, encrypted with your key One app for phone-side models and your desktop local setup. WHAT TO EXPECT Phone models are not the same as huge hosted models in a data center. They are smaller, and that is expected. Performance will vary by hardware. What you gain is control: offline use, private chats on your device or your home machine, and freedom to run the open models you choose. BUILT FOR - People already running LM Studio, Ollama, oMLX or Jan AI who want that stack in their pocket - People who want on-device chat when they are offline or away from Wi-Fi - People who care about keeping conversations off third-party chat servers - Researchers, builders, and privacy-minded users experimenting with local AI INSIDE THE APP - Download and run on-device models on your phone - Connect to LM Studio, Ollama, Jan AI and oMLX - Remote access through LM Mini Connect with your encryption key - Manage and download models on LM Studio from your phone - Local conversation history, folders, and context usage tracking - Personas, plus group chat with up to 5 models in one conversation - Voice chat with Kokoro - Image generation with ComfyUI and Automatic1111 - Web search through SearXNG - Memory so your AI can remember details about you - On-device semantic search across your chats - Background generation, so you don't have to sit and stare at the screen - Themes, backgrounds, and avatars - Import chats from LM Studio - Optional encrypted cloud backups - Custom servers with custom headers - Optional APIs when you want them: OpenAI, OpenRouter, Mistral, DeepSeek, Google Gemini, and more - MCP tool calling - In-app support and feature requests - Dark mode and a chat UI meant for daily use PRIVACY For local use, chats stay on your phone, or between your phone and your own computer. We don't need your conversations to make that core local path work. If you choose optional cloud APIs, those requests go to the provider you pick. SETUP On-device: a compatible phone with enough free storage and memory for the model you pick. Connected mode: LM Studio, Ollama, or oMLX on your computer, plus local network or USB access, or LM Mini Connect for remote use. Install LM Mini and keep your local AI with you. Download LM Mini today and start chatting with your local AI models privately and securely! This app uses Apple Standard EULA license and its available at https://www.apple.com/legal/internet-services/itunes/dev/stdeula/

HAI Chat is a fast, private client for the Hetzner Inference API, the experimental platform serving open AI models. Bring your own API key, pick a model, and chat. WATCH IT THINK The models reason before they answer, and HAI Chat streams that thinking live in a collapsible fold. Choose how much thinking a question deserves, or let Automatic pick a model for each message. OPEN MODELS, HUGE CONTEXT Chat with the open Qwen models Hetzner serves, with context windows of 262k tokens. The model list loads live from the API, so new models appear as Hetzner adds them. TASKS Give HAI Chat a goal and it works through it step by step with your calendar, reminders, contacts, places, the weather and the web, also in the background. Every step shows in the chat, and nothing changes without your OK. ON IPHONE, IPAD, MAC, AND APPLE WATCH Your chats sit in a sidebar next to the conversation, and ⌘K opens any of them. On Apple Watch, ask a quick question from your wrist. REPLIES THAT READ PROPERLY Code, tables, lists, and math formulas are laid out properly, and diagrams are drawn on your device. HTML or SVG code can be previewed with the network shut off. ATTACH WHAT YOU ARE ASKING ABOUT Pictures, text files, PDFs, scans, and web pages. Attach a spreadsheet and get exact sums and charts, worked out on your device. Scan a receipt or a business card and add it to Contacts, Calendar, Reminders, or a spreadsheet with one tap. PROJECTS WITH SOURCES Group chats about one topic in a project with instructions and files. A project can hold hundreds of PDFs, each question gets the passages that match, and the answer names the file and page. WORK WITH TEXT, LEARN WITH CARDS Turn a reply into a document and have parts shortened, rewritten, or translated, with every change shown before you keep it. Make study cards from a PDF or a chat and review them when they are due. PROMPTS, STYLES, AND TOOLS Type / for a ready-made prompt or one you saved, pick a response style, and describe yourself once. Models can check the date, calculate, and convert units on your device. KEEP WHAT MATTERS Search, pin and archive chats, save a reply as Markdown or PDF, and back up all chats to one file, protected with a passphrase if you like. Bring your history over from another assistant's data export. TALK AND LISTEN Dictate on your device, or go hands-free with the camera on and ask about what you point it at. Interrupt an answer by speaking. Use the interpreter for a conversation in two languages, or turn a long text into a briefing that keeps playing with the phone locked. OUTSIDE THE APP Siri, Shortcuts, Spotlight, Handoff, the Action button, and a notification when a reply finishes in the background. Share text, links, and pictures in from any app. PRIVATE BY DESIGN There is no account and no tracking. Your API keys, one per profile, stay in the Keychain on this device, and on your Apple Watch once you send one there, and go only to the Hetzner Inference API. Chats are stored on your device, and nowhere else unless you save a backup or switch on iCloud sync, which puts them in your own private iCloud database that the developer cannot read. Turn on the privacy shield and names, addresses, and numbers are swapped for placeholders on your device before a message leaves it. Temporary chats are never stored. Speech runs on your device. Face ID can lock the app. IN ENGLISH AND GERMAN The interface follows your system language. HONEST BY DESIGN AI answers can be wrong. HAI Chat says so right under the message field. REQUIREMENTS You need your own API key for Hetzner's experimental Inference platform, free of charge while the platform is in its experimental phase. The platform and its model lineup may change or be discontinued at any time. That is part of the experiment. HAI Chat is an independent project. It is not affiliated with, endorsed by, or operated by Hetzner. "Hetzner" is a trademark of Hetzner Online GmbH, used here only to describe compatibility.