A new open source application called Nativ lets users run frontier open AI models directly on their Macs. The app is designed for Apple Silicon (M1 and later) and requires no accounts, subscriptions, or cloud services.
What Nativ offers
Nativ provides a curated library of open models from Google, Cohere, and Liquid AI. It recommends the best model for each user's hardware. Available models include Gemma 4 E2B Instruct from Google (10.28 GB, 128K context), North Mini Code from Cohere (19.38 GB, 500K context), and LFM2.5-VL 1.6B from Liquid AI (3.20 GB, 128K context).
The app includes a chat interface with streaming responses, markdown support, code highlighting, and image input. Every response is generated locally. Users can also view live performance metrics such as tokens per second, memory pressure, thermal state, and time to first token.
Built for Apple Silicon
Nativ is optimized for Apple Silicon using MLX-VLM and is tuned for M-series unified memory and Metal. The developers say there are no wrappers or translation layers. The app supports multiple modalities including language, vision, video, code, and audio. Users can chat with an LLM, caption an image, summarize video, autocomplete code, or transcribe and generate speech.
Integrations with coding agents
Nativ provides a local endpoint that connects to coding agents such as Pi, Codex, Claude Code, Hermes, and OpenCode. Users can keep their existing workflows while running models locally.
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
Open source philosophy
The developers of Nativ have released the entire desktop app under the MIT license. In a manifesto included with the release, they criticize other local AI apps for being proprietary shells built on top of open source engines. They say those apps keep the UI closed, add a paywall, and hide their inner workings.
Nativ is built in the open. Every line of code, every model loader, and every telemetry chart is available for users to read, fork, or submit pull requests. The developers state there is no VC roadmap, no enterprise tier, and no dark patterns that turn prompts into training data. They describe the software as made by researchers and hackers for researchers and hackers.
Free forever
The app is free forever with no accounts, no credits to buy, and no data to sell. Users own the software end to end.
Availability
Nativ is available for download on macOS. The source code is also available for viewing. The latest release is live now.