Ukrainian software developer MacPaw has partnered with Liquid AI to power its products, including the forthcoming AI assistant Eney, with locally hosted models, the company announced on August 5, 2026. The partnership extends to MacPaw's subscription app store SetApp, which serves over 150,000 paying users and is being repositioned around AI applications with credit-based pricing.

Liquid AI, which describes itself as an efficiency-first foundation model company building compute-optimized models for any device, will provide the on-device inference layer. The collaboration includes a system called Elix, along with a local memory architecture, both developed by Liquid AI to run Eney's assistant capabilities without network access. MacPaw says on-device inference eliminates the need for Eney queries to reach the network, allowing the assistant to deliver solutions faster, according to a MacPaw newsroom post published in December 2025.

Liquid AI co-founder and CEO Ramin Hasani said the company selects an architecture tailored to the target hardware before training its models, which allows for efficient on-device intelligence with privacy and security benefits. Hasani also said Liquid AI is building a customization stack around its models that lets them use user data to improve and become more intelligent over time. The degree to which that personalization occurs entirely on-device versus with some network component was not detailed.

MacPaw CEO Oleksandr Kosovan said locally hosted AI models will let users run assistants and agentic workflows offline. Eney, which MacPaw calls a "Computerbeing" — an AI companion designed to think with you, act for you, and evolve around how you work — was unveiled in mid-2025. Since then, MacPaw has shown it as a Mac assistant controllable through natural voice interaction, one that finds where work is happening and handles the connective, attention-surfacing parts of it. Eney and Liquid AI's technology are becoming more integrated across system care, cybersecurity, app discovery, and AI assistance, according to a MacPaw newsroom recap published in July 2026.

For MacPaw, the Liquid AI partnership is not limited to its own products. The company plans to make the local processing architecture available to third-party developers building for SetApp, enabling on-device inference in their applications. MacPaw's planned platform will also provide developers access to cloud models from providers like Google, positioning SetApp as a single entry point for both local and cloud AI capabilities.

MacPaw is experimenting with credit-based pricing for SetApp's AI features. Under this model, users can perform a number of AI operations based on credits, with cost tied to task complexity. The details of how credits map to specific operations were not disclosed. SetApp currently has over 150,000 paying subscribers, and MacPaw plans to focus more heavily on AI apps within the store.

The partnership arrives as the on-device AI inference landscape grows more crowded. Apple, Google, and Microsoft have all invested in local model execution through their respective operating system integrations, and frameworks like ONNX Runtime, Core ML, and llama.cpp have made it increasingly feasible for developers to run small models on commodity hardware. What differentiates MacPaw's approach is the combination of a curated app distribution channel — SetApp's subscription model — with a developer-facing inference platform that spans both local and cloud models.

For developers, the proposition is straightforward: a single API surface for AI calls, with Liquid AI's Elix handling on-device work and cloud providers filling in for tasks that exceed local compute or context limits. Whether MacPaw can attract enough AI-focused developers to SetApp to make that platform meaningful is the open question. The store's existing user base of 150,000 paying subscribers is a credible starting point, but it is modest compared to the distribution of Apple's App Store or even standalone AI tools that have reached tens of millions of users on their own.

MacPaw's broader product portfolio spans system utilities (CleanMyMac, CleanMyPhone), VPN software (ClearVPN), and cybersecurity through its Moonlock Engine division. Eney appears designed to serve as the connective layer across these products — an assistant that can orchestrate system maintenance, surface security concerns, and handle general Mac productivity tasks. The Liquid AI partnership gives that vision a technical foundation: local inference for latency-sensitive and privacy-critical operations, with the option to escalate to cloud models when needed.

Liquid AI's involvement also signals continued momentum for the thesis that purpose-built, architecture-selected models can outperform general-purpose models on edge hardware. Hasani's claim that Liquid AI tailors architecture to hardware before training is a meaningful departure from the dominant approach of training large general models and then compressing them for deployment. If Elix delivers efficient inference on Mac hardware without unacceptable quality tradeoffs, it would validate that approach in a consumer-facing context.

MacPaw did not provide a specific launch date for the Eney local engine or the developer-facing inference platform. The company's newsroom landing page, as of this reporting, did not list a dated press release about the Liquid AI partnership among its visible headlines.