MacPaw Partners With Liquid AI on Mac AI Stack
MacPaw announced a long-term partnership with Liquid AI on August 5 to co-develop an on-device AI stack for macOS, beginning with its Eney assistant. The companies will combine Liquid Foundation Models with MacPaw's Elix inference and Mnemos memory technologies, according to their announcement. 9to5Mac reports that MacPaw also intends to extend the shared stack to additional products and potentially developers through Setapp.
MacPaw announced a long-term partnership with Liquid AI on August 5 to co-develop a local AI technology stack for macOS. According to the companies' PRNewswire announcement, the initial application will be Eney, MacPaw's macOS assistant, with results expected later this year.
The collaboration combines Liquid AI's Liquid Foundation Models (LFMs) with MacPaw's on-device components: Elix for inference and Mnemos for memory. The announcement describes LFMs as being developed and fine-tuned for Eney's tasks, with core workloads running locally through Elix on Apple silicon. It also states that cloud models will remain available for tasks where they are more suitable.
MacPaw CEO and founder Oleksandr Kosovan said the companies are building a stack intended to keep personal data on-device while using cloud resources when appropriate. Liquid AI's announcement also identifies Mercedes-Benz, Insilico Medicine, and Shopify as existing users of its models for on-device intelligence.
Extending beyond the assistant
9to5Mac reports that Eney keeps reasoning, contextual search, skill execution, and conversation history on-device when possible. The publication characterizes the new work as a jointly developed underlying stack rather than a conventional integration of a third-party model.
According to 9to5Mac, MacPaw intends to bring the technology to more products after testing and implementation in Eney, and to make the shared AI stack available to Mac developers through Setapp. The report describes a potential model in which participating apps expose capabilities and contextual information to Eney, allowing the assistant to provide answers or perform relevant actions within those apps.
What local execution changes
For ML practitioners, the technical emphasis is on a familiar hybrid inference pattern: compact local models and device-resident state for latency-sensitive or privacy-sensitive interactions, with cloud models retained for workloads requiring greater capability. That architecture can reduce network dependence, but comparable cross-application assistants also require carefully scoped permissions, reliable tool interfaces, and clear boundaries around which app context is available to a model.
The announcement does not provide model sizes, benchmark results, supported hardware requirements, or details on Eney's cross-application permission model. Those implementation details will determine how broadly the stack can operate across Apple's installed Mac base and whether developers can integrate it without creating fragmented tool or context schemas.
Key Points
- 1MacPaw and Liquid AI are combining LFMs, local inference, and persistent memory, with Eney identified as the first product target.
- 2The proposed design retains cloud-model access while shifting core assistant tasks to Apple silicon, creating a hybrid local-and-cloud inference architecture.
- 3Comparable cross-app assistants depend on permission boundaries and tool interfaces as much as model quality, especially when accessing application context.
Scoring Rationale
The partnership is a notable on-device AI product initiative aimed at the Mac ecosystem, but it is an announced development effort rather than a released platform. Its practitioner relevance depends on later technical disclosures around models, integrations, permissions, and developer access through Setapp.
Sources
Primary source and supporting public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
