MacPaw and Liquid AI Partner to Build On-Device AI Stack for macOS
The collaboration introduces Elix and Mnemos to bring high-performance local inference to Mac apps.
MacPaw has entered a long-term strategic partnership with Liquid AI to develop a dedicated on-device AI technology stack for macOS. The move aims to shift AI processing from the cloud to local hardware, prioritizing user privacy and reducing latency.
The collaboration centers on two primary components: Elix, a system for on-device inference, and Mnemos, a local memory system. According to reports from TechCrunch and 9to5Mac, this infrastructure is designed to allow AI models to run natively on Mac hardware. The first application to implement this technology will be Eney, MacPaw's AI assistant for macOS. To demonstrate the potential of this approach, Liquid AI's LFM2.5-2.6B model has shown the ability to achieve decode speeds of 220 tokens per second on an Apple M5 Max chip.
The Shift to Edge AI
MacPaw, a Ukrainian software company recognized for its suite of Mac utilities, is pivoting toward AI-integrated applications. This transition reflects a broader industry movement toward "Edge AI," where large language models (LLMs) operate on consumer devices rather than remote servers. By utilizing Liquid AI's Liquid Foundation Models (LFMs), which are optimized for compute efficiency on edge devices, MacPaw can lower operational costs while offering a more responsive user experience.
Ecosystem Implications
This partnership extends beyond MacPaw's own internal tools. The company intends to make this on-device inference layer available to third-party developers building for Setapp, MacPaw's subscription-based app store. By providing a standardized AI layer, MacPaw is attempting to cultivate an AI-centric ecosystem for macOS. This could position Setapp as a primary hub for high-performance, privacy-preserving AI applications, giving developers the tools to implement complex AI features without requiring their own massive cloud infrastructure.
What to Watch
As the rollout begins with Eney, the industry will be watching to see how effectively third-party Setapp developers adopt the Elix and Mnemos systems. The success of the partnership depends on whether these local models can maintain the utility of cloud-based LLMs while operating within the hardware constraints of a laptop. Future updates will likely focus on the breadth of model support within the Elix framework and the actual performance gains realized by end-users across different Mac hardware generations.