Nativ: Run frontier open models locally on your Mac
Points and comments are a snapshot, not live.
Nativ is an open source Mac app to run AI models locally on Apple Silicon.
Nativ (v0.0.1, MIT licensed) runs frontier open models on Apple Silicon (M1+) with no cloud, accounts, or subscriptions. It provides a chat interface with streaming responses, telemetry (tokens/sec, memory pressure), and supports multiple modalities including text, image, video, audio, and code. Built on MLX-VLM and optimized for M-series unified memory, it integrates with coding agents like Pi, Codex, and Claude Code via a local endpoint. The project is fully open source, with the developer arguing that other local AI apps keep their UIs closed and behind paywalls.
What commenters are saying
Several commenters welcome another competitor to Ollama and LM Studio, noting this app's pedigree from the mlx-vlm maintainer. Some question the use of the term "frontier models," arguing it misleadingly implies top-tier models like Fable rather than the smaller Gemma 4 and similar models that can run on a Mac. Others defend "frontier" as referring to the Pareto frontier of capability relative to size and cost. One comment recommends oMLX as a more complete current alternative, while another reports papercuts and bugs with oMLX and is willing to try Nativ despite its early stage. A few users point to additional local MLX apps: rapid-mlx, mtplx.com, and jan.ai.