Local LLM vs Hermes Redde

Local LLM runs language models on your iPhone. No account is required. Chat, ask about photos, work with documents and use live voice. With Canvas, build websites, interactive pages and games from a conversation, then preview and edit them in the app. Features depend on the selected model. Download models from the catalog, or use Apple’s on-device model on supported devices when available. Core AI models require iOS 27 or later. Other supported models remain available on older systems. Inference and conversation history stay on your device. Model downloads, the catalog and model artwork use the network. Optional Web Search sends a query derived from your message to Tavily and may read result websites. Weather requests use Apple geocoding and WeatherKit for the requested city, without requesting device location. Voice input prefers on-device recognition. If it is unavailable or fails, Apple’s online speech service may process audio when network speech recognition is enabled. You can disable this in Privacy settings. Requires iOS 17 or later. Model availability depends on your device, OS, memory and storage. Downloads require internet; supported offline chat works after setup. Models can make mistakes. Do not rely on replies for medical, legal or financial decisions.

Redde is a complete iPhone, iPad and Mac client for Hermes, the open-source AI agent, with CarPlay and an Apple Watch app. Talk or type, watch it work, let it get ahead of your day, and give it a crew: bots with faces and voices of their own. You bring the backend. Redde has no account and no analytics, and your conversations go only to your server, and to a voice service if you pick one. AHEAD OF YOUR DAY • Morning Ideas, a Week Ahead and an Evening Wrap-up arrive as cards to swipe. Before an event, a heads-up says what to bring, and Leave-By Alerts work out from your calendar and Apple Maps when to go. • "Let me know when it ships", "check back Friday": watches and follow-ups come back on their own. Places tell your agent you've arrived; Health-aware days, if you allow. • Answers come as cards: your day as a timeline, the weather by the hour, places on a map, an email to check before it sends, a list to tick off. Approvals say in plain words what will happen. BOTS • Each Hermes profile is a bot with a face, a voice and a chat of its own. Start from a template (Researcher, Writer, Coder, Home, Planner) and make its job, model and tools your own. • Type @ to hand work to another bot, or put two to six in a group chat. Give one a schedule and its answer arrives with a notification, even with Redde closed. TALK, DON'T TYPE • Tap the mic or say "Hey" and your agent's name, ask, hear the answer, or show it what the camera sees. Speech is recognized on your device; replies are read by the built-in voice, Kokoro, ElevenLabs or your Hermes. • Or talk live, interrupting any time, with GPT-Live, Gemini Live, Grok Voice or ElevenLabs on your own key. An agent you allow can call you. • Hands-free until you say "that's all". CarPlay, Apple Watch, Siri, widgets and the Action Button. Spoken prefixes switch models or go to a bot: "Opus, review this diff". SEE THE AGENT WORK • Reasoning and tool calls live, with a Live Activity; lock the phone and it carries on. • Approve or answer from a notification, out loud, or with a nod in AirPods. Notifications reach you with Redde closed, encrypted end to end. A FULL CLIENT • Chats live in your gateway's session ledger: start on the phone, continue on the desktop. Search, pin, fork, archive, or go incognito. • Projects, scheduled jobs, the Kanban board, skills, plugins, memory, MCP servers, files and an SSH terminal. • Phone tools: your agent can use your calendar, reminders, contacts, photos and files, set alarms and start calls and texts, each only if you allow it. • Markdown, code, tables, diagrams and math; pictures, PDFs and audio open in place. • Attach photos, files or a day's calendar, or share from any app. Regenerate, edit and resend, or switch models with /model. YOUR BACKEND, YOUR RULES • Connect through the Hermes Dashboard or the Hermes API, on your LAN, over a tailnet, or behind Cloudflare Access, Pangolin or a reverse proxy. Several servers in one list, synced with iCloud. Sign in with a password or a browser. • Or go straight to any OpenAI-compatible endpoint: llama.cpp, llama-swap, vLLM, Ollama or a hosted provider. • Set up by QR code or link. Credentials stay in the Keychain; a key your agent asks for goes sealed straight to your server. • In 17 languages. Face ID lock, seven themes, iPad split view, iPhone Duo, and a native Mac app. • The one service we run is an optional relay for notifications, which can't read them. REQUIRES A Hermes agent (github.com/NousResearch/hermes-agent) or an OpenAI-compatible endpoint; bots need Hermes. Redde includes no model and no AI service of its own. Hermes is an open-source project by Nous Research. Redde is an independent client, not affiliated with it.