MacPaw and Liquid AI Build a Private On-Device AI Stack Challenging Apple Intelligence
Liquid AI and MacPaw are co-building a full on-device AI stack for Mac, with LFMs, local inference, and persistent memory — opening up to thousands of Setapp developers
- MacPaw and Liquid AI partner to co-build a full on-device AI stack for macOS, targeting millions of Mac users. Read announcement
- Three-layer stack: Liquid Foundation Models (LFMs) + Elix (local inference engine) + Mnemos (persistent memory layer) — all running on Apple silicon.
- Eney, MacPaw's macOS AI assistant, is the first product on the stack; production release expected later this year.
- Setapp is the real endgame — the same stack could be opened to thousands of Mac developers via MacPaw's 150,000+ user app marketplace. TechCrunch coverage
- Liquid AI is an MIT spinout valued at ~$2B, with $297M raised; LFMs run at 220 tokens/sec on Apple M5 Max hardware.
- No financial terms disclosed; this is a co-development agreement, not an acquisition or equity deal.
MacPaw and Liquid AI have announced a strategic, long-term partnership to co-develop a full local AI stack for the Mac. The goal is straightforward: bring private, fast, offline-capable AI to millions of Mac users , without routing every query through a cloud server. The first product built on this stack is Eney, MacPaw's macOS AI assistant, with a production release planned for later this year.
Three layers, one stack
The collaboration combines Liquid Foundation Models (LFMs), optimized for macOS, with MacPaw's Elix for inference and Mnemos for memory. Think of it as a three-layer sandwich:
- LFMs , Liquid AI's foundation models, fine-tuned specifically for macOS assistant tasks and designed to run efficiently on Apple silicon.
- Elix (Eney Local Intelligence MLX) , MacPaw's local engine that automatically defaults to local processing and storage, meaning reasoning, context search, skill execution, and conversation history are kept on the device rather than being sent to the cloud.
- Mnemos , MacPaw's context system that continuously indexes your macOS environment, providing highly personalized and situationally appropriate context.
MacPaw's CEO Oleksandr Kosovan said locally hosted AI models will also give users the ability to run assistants and agentic workflows offline. The goal is AI that keeps personal data private, responds quickly, and operates offline.
Why this is different from Apple Intelligence
Notably, Apple already provides its own local models to developers. So what's the angle here? Liquid AI CEO Ramin Hasani put it plainly: "Before training our models, we select an architecture that is different and tailored to the hardware. That allows us to really have the most efficient version of intelligence that runs directly on the device, with benefits like privacy and security."
The deeper pitch is adaptability. Hasani told TechCrunch: "We are also building a customization stack around models. This means that with user input, the models can use the data and improve. We want our models to be adaptable and become more intelligent over time." That's a meaningful distinction , Apple Intelligence models don't learn from your personal usage patterns. Mnemos is specifically designed to do that.