MacPaw today has announced a new partnership with Liquid AI – a development focused on bringing native AI experiences to Mac devices. The plan is to have the native AI functions come to SetApp and Eney, MacPaw’s native AI assistant for macOS, later this year. Appleosophy has covered MacPaw’s developments before, as a Ukraine-based software development and research company, this newly announced partnership is sure to bring interesting updates to their existing suite of applications on Mac.
Who are Liquid AI?
Liquid AI is a foundation model company spun out of MIT CSAIL. Its Liquid Foundation Models (LFMs) are general-purpose AI models built on an architecture designed for efficiency, created to run reliably on real devices as opposed to in data centers. Liquid AI’s models already run on device for customers including Mercedes-Benz, Insilico Medicine, and Shopify.
Who’s doing what?
MacPaw Research continues building Elix, Mnemos, and the agent framework behind Eney, with a research practice that spans the range from fundamental research to engineering; available at research.macpaw.com/publications, with more to come after the release. The partnership accelerates that work, enabling faster development cycles and increased innovation.
The reported partnership is joint a R&D initiative between two engineering teams. Liquid AI develops and adapts unique, specific, efficient on-device LFM models, built for Eney’s tasks, Mac assistance and memory, capable of tool calling, memorizing, and more; their goal is advanced models, leading the market on metrics. MacPaw brings almost two decades of system-level macOS engineering and its own technologies, Elix and Mnemos. MacPaw’s engineers take part in training and evaluating the models directly, so improvements are driven by the tasks Mac users actually bring to the assistant.
A current development that MacPaw is working on is dynamic routing based on type of request and available processing cloud or on-device within their AI tool, Eney. At the time of publication, Eney is currently cloud only, with the expectation that models can be run on device with the update later this year – hopefully as a deliverable from the MacPaw & Liquid AI partnership.
Other Optimizations are also being made for M-series Apple Silicon chips to utilize the full power of the chip and different AI models for a range of tasks. Eney remains as a permission-based tool, so sensitive areas, files or content are shared on an approved basis. User privacy has been a contencious point for a lot of AI software development, and the Setapp developers are mindful that many users will hold personal data which they don’t want shared with AI services.
MacPaw welcomes everything that advances on-device AI on the platform. MacPaw developers have tested Apple Intelligence directly, and found that Apple’s foundation models (AFM) on their own aren’t enough for what Eney assistant needs to do — according to the developers, building genuinely capable on-device assistance isn’t possible with AFM alone (at present). That’s exactly why we’re building the full stack for our own products and ecosystem: models adapted to our assistant’s real tasks, an inference framework, and a memory layer working together. The partnership with Liquid AI is what makes that possible and how to actually build it.
Appleosophy is excited to see how this partnership can flourish and provide a new level of integration to the macOS AI race – as companies work to having native models available on Macs worldwide!
