MacPaw Partners with Liquid AI to Provide On-Device Inference for Developers in Its App Store

The Ukraine-based app developer MacPaw has collaborated with Liquid AI to enhance its products with locally hosted AI models, aiming to eventually provide this tech stack to developers. The company is also getting its SetApp app store ready for AI applications, offering credit-based plans for users.

MacPaw is in the process of developing its AI assistant Eney, which was introduced last year, and is now aiming to create a locally hosted version. For this purpose, it has engaged Liquid AI to create an on-device inference system known as Elix, along with a local memory system.

“We choose an architecture that is specifically tailored to the hardware before training our models. This enables us to create the most efficient version of intelligence that operates directly on the device, offering benefits like enhanced privacy and security,” stated Ramin Hasani, co-founder and CEO of Liquid AI, in an interview with TechCrunch.

Oleksandr Kosovan, CEO of MacPaw, mentioned that locally hosted AI models will empower users to operate assistants and agentic workflows even offline.

Interestingly, Apple already provides its own local models for developers. However, Hasani insists that Liquid AI’s models prioritize performance for various capabilities.

“We are also creating a customization stack for our models. This means that with user feedback, the models can utilize data to enhance themselves. Our goal is for the models to be flexible and evolve in intelligence over time,” Hasani explained.

MacPaw intends to emphasize AI applications within its subscription-based app store, SetApp, which currently boasts over 150,000 paying users.

Once the local processing architecture with Liquid AI is established, MacPaw aims to provide this technology to developers, enabling on-device inference in their applications. Kosovan mentioned that the platform will also grant access to other cloud models from companies like Google, serving as a comprehensive resource for developers.

MacPaw is already testing a credit-based pricing model for its app store, where users can execute a certain number of AI operations depending on their credits and the complexity of the task.