Analysis
MacPaw is partnering with Liquid AI to give developers building for its Setapp app store on-device AI inference, letting apps run models directly on a user's machine instead of routing every request through a cloud API, according to [TechCrunch](https://techcrunch.com/2026/08/05/macpaw-taps-liquid-ai-to-offer-on-device-inference-to-devs-building-for-its-app-store/).
Why On-Device Matters Right Now
Liquid AI has built its identity around an efficient, non-transformer 'liquid neural network' architecture specifically designed to run well on local hardware rather than requiring large cloud GPU clusters -- a genuine technical differentiator against most foundation model providers whose products assume a cloud API call by default. On-device inference sidesteps per-token cloud costs and network latency entirely, addressing exactly the kind of unpredictable AI spending problem that pushed Microsoft to cap internal token budgets this same week.
Platforms Building AI Infrastructure Into Developer Tooling
For MacPaw, baking this partnership directly into Setapp's developer tooling is a meaningful platform move -- rather than leaving each individual developer to separately negotiate cloud AI vendor relationships and manage their own inference costs, the app store itself is absorbing that infrastructure decision on developers' behalf.
What to watch: whether other app-store and developer-platform operators follow with similar on-device inference partnerships as unpredictable cloud AI costs become a bigger concern across the industry, and whether Liquid AI's architecture holds a real efficiency edge over increasingly optimized cloud-based small models.
