Nativ: Run frontier AI models locally on Mac
Original: Nativ: Run frontier open models locally on your Mac
Why This Matters
Fully open-source local AI tooling on Apple Silicon signals growing demand for private, subscription-free on-device inference.
Nativ is a 100% open-source, MIT-licensed desktop app for macOS that lets users run frontier AI models locally on Apple Silicon (M1+) — no accounts, no subscriptions, and no cloud required. It supports models from Google, Cohere, and Liquid AI.
Nativ is a free, fully open-source macOS application built for Apple Silicon that enables local execution of frontier open AI models. It is built on MLX-VLM and optimized for M-series unified memory and Metal GPU, with no wrappers or translation layers. The app features a curated model library including Google's Gemma 4 E2B (10.28 GB, 128K context, vision + audio), Cohere's North Mini Code (19.38 GB, 500K context, code + tools), and Liquid AI's LFM2.5-VL 1.6B (3.20 GB, 128K context, vision + language). Nativ supports multiple modalities — text, image, video, code, and audio — and provides a clean chat interface with streaming responses, markdown rendering, and image input. A telemetry panel shows live tokens/sec, memory pressure, thermal state, and time-to-first-token. The app also exposes a local endpoint compatible with coding agents such as Claude Code, Codex, Hermes, and OpenCode. The project is MIT licensed, has no enterprise tier, and explicitly states it does not use prompts as training data. Source code is fully available on GitHub.