Ukrainian software company MacPaw and Cambridge-based foundation model startup Liquid AI have announced a long-term strategic partnership to co-develop a local AI technology stack for Mac computers. The deal pairs Liquid AI's on-device foundation models with MacPaw's own inference engine, called Elix, and a persistent memory system called Mnemos. MacPaw's AI assistant Eney will be the first product to ship with this technology, with results expected later this year.

The companies say their joint goal is to bring private, offline-capable AI directly to Apple silicon hardware. MacPaw also plans to open the local inference stack to third-party developers through its SetApp software marketplace.

What the Partnership Actually Builds

The collaboration has two technical components running in parallel.

First, Liquid AI will develop and fine-tune its Liquid Foundation Models specifically for the tasks Eney performs. Those models will run locally through Elix, MacPaw's on-device inference framework optimized for Apple silicon. Cloud-based models will remain available as a fallback when a task demands more compute than local hardware can deliver.

Second, the two companies are co-developing Mnemos, MacPaw's memory layer. Mnemos lets the assistant retain context across separate interactions, so it can recall prior conversations and improve its responses over time without sending personal data to a remote server.

"We believe intelligence should live where people work: private by design, fast by default, and be able to reach the cloud when that's the better tool," said Oleksandr Kosovan, CEO and founder of MacPaw.

According to Liquid AI's co-founder and CEO Ramin Hasani, the company tailors its model architecture to specific hardware before training begins. "Before training our models, we select an architecture that is different and tailored to the hardware. That allows us to have the most efficient version of intelligence that runs directly on the device, with benefits like privacy and security," Hasani told TechCrunch.

Who Is Liquid AI

Liquid AI was founded in 2023 as a spinout from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL). Its four co-founders are Ramin Hasani, Mathias Lechner, Alexander Amini, and Daniela Rus, who serves as director of MIT CSAIL. The company is headquartered in Cambridge, Massachusetts.

The startup raised a $37.6 million seed round in late 2023, then closed a $250 million Series A led by AMD in December 2024. That round valued the company at over $2 billion, and AMD's involvement reflects the chipmaker's interest in Liquid AI's hardware-efficient model design.

Liquid AI's core technical contribution is its Liquid Foundation Models (LFMs), which use a non-transformer architecture rooted in dynamical systems and liquid neural networks. The approach was developed through Hasani's doctoral research at TU Wien and his subsequent postdoctoral work at MIT. Unlike conventional transformer-based models that keep all parameters active during inference, LFMs are designed to activate parameters selectively based on task complexity. The result is lower memory usage and faster inference on constrained hardware like phones and laptops.

The company's LFM2 family, released in mid-2025, includes models ranging from 350 million to 3 billion parameters. Liquid AI claims these models deliver twice the decode and prefill performance of comparable open-source alternatives like Qwen3 when running on CPU. MacPaw and Liquid AI are not the startup's first enterprise deployment. Liquid AI already supplies models to Mercedes-Benz, Shopify, and Insilico Medicine.

MacPaw's Pivot From Utilities to AI Platform

MacPaw was founded in 2008 in Kyiv by Oleksandr Kosovan, who was a university student at the time. The company built its reputation on CleanMyMac, a system optimization utility that now runs on roughly one in five Macs worldwide. MacPaw serves over 30 million users globally, employs more than 500 people, and maintains offices in Kyiv and Boston.

But the company has been steering toward AI since September 2025, when Kosovan publicly declared a pivot to an AI-first business. Volodymyr Kubytskyi, who became MacPaw's AI director in August 2024, authored the company's AI strategy for 2025 and 2026. Under that strategy, each MacPaw product now has its own dedicated AI team rather than relying on a single centralized group.

The centerpiece of this transformation is Eney, which MacPaw calls a "Computerbeing." It launched in closed beta on SetApp in mid-2025 and was demonstrated at CES 2025 and CES 2026. Eney sits on the edge of the Mac display, accessible by moving the cursor toward it, and responds to voice or text commands. It can execute tasks, manage files, summarize content, and integrate with third-party apps through expandable skills.

At CES 2026, MacPaw also announced a separate partnership with Respeecher, a Ukrainian voice synthesis company whose technology has appeared in dozens of Hollywood productions. That deal added natural-sounding text-to-speech capabilities to Eney.

The Liquid AI partnership is the infrastructure layer beneath all of this. Instead of relying on cloud APIs for its core intelligence, MacPaw wants Eney's primary reasoning to happen on the user's machine.

SetApp as an AI App Distribution Channel

MacPaw also intends to use this partnership to transform SetApp, its subscription-based software marketplace. SetApp launched in 2017 and currently catalogs more than 240 Mac applications. It has over 150,000 paying subscribers.

Kosovan's vision is to position SetApp as a one-stop distribution point for AI-powered apps. Once the local inference stack is stable, MacPaw plans to make Elix and the Liquid Foundation Models available to third-party developers building for the platform. Developers would be able to run on-device inference within their own apps, and the platform would also provide access to cloud models from providers like Google.

MacPaw is already experimenting with a credit-based pricing model for AI operations within SetApp. Under this system, users would receive a certain number of credits with their subscription, and each AI task would consume credits based on its computational complexity. This mirrors how API-based pricing works in the developer world but packages it for consumer use.

The company shut down SetApp Mobile, its alternative iOS app store for EU users, in February 2026 after concluding that Apple's Digital Markets Act compliance terms did not align with SetApp's business model. The Mac version continues to operate and will be the primary vehicle for the AI developer platform.

Competing With Apple on Apple's Own Hardware

The most obvious competitive tension in this announcement is with Apple itself. At WWDC 2026 in June, Apple unveiled Core AI, a new framework that replaces the nine-year-old Core ML as the primary tool for developers running AI workloads on Apple hardware. Core AI is specifically designed for generative AI tasks and supports both Apple's own foundation models and third-party models running on Apple silicon.

Apple's AFM (Apple Foundation Model) family is now in its third generation, and the company offers developers direct access to its on-device large language model through the Foundation Models framework.

MacPaw and Liquid AI are essentially arguing they can deliver a better on-device AI experience than the platform owner. Hasani's position is that Liquid AI's hardware-specific model architecture provides performance advantages that generic models cannot match. He also emphasized a customization angle: "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."

Whether that pitch holds up against Apple's deep integration with its own silicon remains to be seen. Apple controls the hardware, the operating system, the Neural Engine, and the development toolchain. A third-party stack running on that same hardware will need to show measurable performance gains to win developer adoption over the path of least resistance.

The Ukrainian Tech Connection

Both MacPaw and Respeecher are Ukrainian companies, and MacPaw's continued growth during wartime has drawn attention. Despite Russia's full-scale invasion beginning in 2022, MacPaw's staff grew 22% that year and another 14% the following year, bringing its workforce past 540. The company has maintained its Kyiv headquarters throughout the conflict.

MacPaw Research, the company's dedicated research arm, focuses on LLM efficiency, AI memory, and human-computer interaction. It maintains academic partnerships with multiple Ukrainian universities and has published peer-reviewed work on efficient on-device inference.

Kosovan is also a co-founder of SMRK Venture Capital Fund, which provides venture investment to Ukrainian IT startups.

What Comes Next

MacPaw says the on-device AI stack powered by Liquid Foundation Models will appear in Eney first, with results expected before the end of 2026. If the infrastructure proves stable, the company plans to package it as shared tooling for its full product ecosystem and eventually offer it to external developers through SetApp.

Liquid AI, for its part, announced an updated LFM2.5 model series in 2026 and continues to expand its enterprise partnerships. The MacPaw deal adds consumer software to a client roster that has so far leaned toward industrial and e-commerce deployments.

The credit-based pricing model for AI operations on SetApp has not been finalized, and MacPaw has not disclosed specific pricing tiers or launch timing for the developer-facing infrastructure.