MacPaw is making a major play in the on-device AI space, partnering with Liquid AI to bring local inference capabilities to developers building apps for its platform. The Ukrainian software maker - best known for CleanMyMac - is rebuilding its AI assistant Eney to run entirely on-device using Liquid AI's efficient models, signaling a broader shift away from cloud-dependent AI tools. For developers in MacPaw's ecosystem, this means access to privacy-preserving AI that runs without sending data to remote servers.
MacPaw just landed a partnership that could reshape how Mac developers think about AI integration. The company's teaming up with Liquid AI, the MIT spinout that's been turning heads with its efficient neural network architectures, to bring on-device inference to its app store ecosystem.
At the center of this partnership is Eney, MacPaw's AI assistant that's getting a complete overhaul. Instead of pinging cloud servers for every request, the new version will run Liquid AI's models directly on users' Macs. It's a technical shift with major implications - no internet dependency, no data leaving the device, and no per-query costs eating into margins.
The timing couldn't be better. Apple's been pushing on-device AI hard with its Neural Engine chips, and developers are hungry for tools that take advantage of that silicon without the headaches of deploying their own models. MacPaw's essentially offering a shortcut - drop Liquid AI's inference into your app through their platform, and you're running local AI without building the infrastructure yourself.
Liquid AI's models are particularly well-suited for this. The startup emerged from MIT with a focus on "liquid neural networks" - architectures designed to be more efficient and adaptable than traditional transformers. That efficiency translates directly to battery life and performance when you're running inference on a laptop instead of a server farm. While OpenAI and Anthropic are locked in an arms race for larger models, Liquid AI's been optimizing for the constraints of edge devices.
For MacPaw, this partnership extends beyond just Eney. The company's building out an alternative app distribution platform for Mac, positioning itself as a developer-friendly alternative to Apple's App Store. Offering built-in AI inference capabilities gives developers a compelling reason to build for MacPaw's platform - it's infrastructure they'd otherwise have to cobble together themselves or pay cloud providers for.
The broader context here is the rapid shift toward edge AI. Companies from Meta to Microsoft are racing to get their models running locally, driven by privacy regulations, latency requirements, and the sheer cost of cloud inference at scale. MacPaw's betting that developers building for Mac want those same benefits without the complexity of optimizing models for Apple Silicon themselves.
There's also a competitive angle. Apple's own AI features run on-device, but they're tightly controlled and not easily accessible to third-party developers. MacPaw's essentially creating an alternative AI layer for the Mac ecosystem - one where developers have more control and users get more choice about which AI tools they use.
The partnership raises interesting questions about the future of AI deployment. If MacPaw can successfully offer plug-and-play on-device AI to its developer community, it could pressure other platforms to follow suit. The days of every app needing its own cloud inference backend might be numbered, replaced by platform-level AI services that run locally.
Liquid AI gets distribution out of this deal - a pathway to reach developers building for millions of Mac users. For a startup competing against well-funded giants, that's valuable real estate. The company's raised significant funding but still needs to prove its models can compete in real-world applications. MacPaw's platform offers exactly that testing ground.
Privacy advocates will watch this closely. On-device AI solves a lot of the thorny data protection issues that plague cloud-based assistants. If MacPaw can deliver a compelling user experience with Eney while keeping everything local, it becomes a proof point for privacy-preserving AI that actually works.
The technical implementation details matter here. Running inference on-device means optimizing for varied hardware - not every Mac has the same Neural Engine capabilities. Liquid AI's models need to gracefully scale across M1, M2, M3 chips and beyond, delivering consistent experiences without draining batteries or hogging memory. That's a harder problem than just throwing more GPU power at it in the cloud.
For developers evaluating MacPaw's platform, the calculation is straightforward - build AI features without managing cloud infrastructure, paying per-token costs, or worrying about data leaving user devices. The tradeoff is betting on MacPaw's distribution platform and Liquid AI's model performance versus going with established cloud providers.
MacPaw's partnership with Liquid AI represents a significant bet on the future of edge computing and privacy-focused AI. By offering developers ready-made on-device inference through its platform, the company's not just upgrading its own assistant - it's building infrastructure that could define how Mac apps integrate AI going forward. If the technical execution delivers on the promise, this could accelerate the shift away from cloud-dependent AI tools toward local models that respect user privacy while delivering comparable performance. The real test comes when developers start building with these tools and users experience whether on-device AI can truly match the capabilities they've come to expect from cloud services.
