CoreAI Zoo vs Marshal Stack

CoreAI Zoo puts the open-weight AI model zoo on your iPhone, iPad and Mac. Pick a model, download it once, and run it entirely on your device. No account, no server, and nothing you type, say or photograph ever leaves the device. CHAT Stream answers from open language models such as Qwen3, Gemma, LFM2.5, Granite and Mistral, from 0.6B to 30B+ parameters depending on your device's memory. Vision-language models (Qwen3-VL, MiniCPM-V, LFM2.5-VL) answer questions about your photos. VISION Ask questions about an image, read text with on-device OCR, and describe what the camera sees. AUDIO Transcribe speech with Whisper large-v3-turbo, Parakeet and Qwen3-ASR. Talk to an audio-understanding model, generate speech with Kokoro and VoxCPM, and try on-device music generation. CAMERA Live depth estimation, object detection (YOLOX, RF-DETR), action recognition and super-resolution on the camera feed. TOOLS Experimental models such as diffusion-based language models. BENCH Measure load time, tokens per second and memory for any chat model on your own hardware. HOW IT WORKS Every model runs through Core ML on Apple silicon. Weights are downloaded from the Hugging Face Hub on first use over HTTPS, then cached, so the app works offline afterwards. One model is resident at a time to stay within memory limits. New models arrive through a live catalog without an app update. GETTING STARTED Start with Qwen3-0.6B (454 MB) for the fastest first run. Larger models state the free space and memory they need; a recent iPhone Pro or an Apple silicon Mac runs the mid-size ones comfortably. Model weights follow their own licenses, listed on each model's Hugging Face page. The app itself is BSD-3-Clause and part of the open CoreAI model zoo. Report issues and request models at github.com/john-rocky/coreai-model-zoo.

Marshal Stack is one screen for everything you have already shipped. Your apps in the App Store, the sites and APIs behind them, the servers you run yourself and the coding agents working on your projects — all read by one rule: green is fine, amber needs a look, red is down. WHAT IT WATCHES • App Store — rating, review count, latest version, and the release state: in review, rejected, live. • Sites and APIs — checked every minute, with the latency next to the result. A notification arrives when an address stops answering twice in a row, and again when it recovers. • Your own servers — CPU, memory, disk, load and the heaviest processes, reported by a small agent you install with one line on a machine you already own. • Cloud servers — connect a read-only Hetzner Cloud token and see your servers' status, CPU, disk and network next to the rest. • Alerts that say what to do, not only what broke — memory over the threshold, a process running from /tmp, a port that was not open yesterday. WAITING FOR YOU When an agent working on your project needs a decision, the question appears on Home and in a notification. Answer with one tap or a typed reply; risky actions ask you to unlock and confirm. The first answer wins everywhere. AGENT USAGE See how much your Claude Code agents used: the 5-hour and weekly usage limit, when it resets, and tokens by task, project, role and model over 7 or 30 days. YOUR KEYS, YOUR MACHINES Connect your own App Store Connect key, Hetzner token and GitHub repositories, or none at all. The server agent dials out over HTTPS: it opens no port, reads no files, touches no databases, and uninstalls with one command. There is no registration form. Sign in with GitHub, or just open the app and look around on sample data bundled inside it — no account, no keys, no network. ON THE WRIST The same three states on Apple Watch, plus your projects ranked worst first. A glance tells you whether anything needs you, and a tap opens the one thing that does. THIRD-PARTY SERVICES Marshal Stack works with services you connect yourself: GitHub (sign-in and repositories), App Store Connect (app ratings and release state), Hetzner Cloud (read-only server inventory), Trello and Telegram (optional task boards and alerts), and Claude Code (the coding agents whose questions and usage you see). Marshal Stack is an independent app and is not affiliated with, endorsed or sponsored by GitHub, Hetzner, Trello, Telegram or Anthropic. No ads, no third-party trackers.