Chinese Startup Om AI Develops Multimodal Models for Edge Devices

As competition in the large model space shifts from parameter scale to real-world applicability, Chinese companies specializing in edge solutions are gaining more attention. One such player is Om AI Technology, which has focused on developing versatile multimodal models for edge devices since its founding in 2021. Unlike many competitors, the company does not aim to create massive cloud models but instead focuses on compact solutions for local devices.

At the BEYOND Expo 2026, Om AI introduced OttoBox AI Studio—a tool for content creation based on local computing. The solution targets media professionals and content creators, offering video analytics, media asset selection, script generation, and fast editing. According to developers, the product is designed to assist in the AI-first content era, enhancing creative workflow efficiency.

Om AI’s key technological advantage lies in developing models for video understanding with minimal computational overhead. Unlike traditional approaches requiring powerful cloud GPUs, the company uses compact edge models capable of running on local devices. This reduces inference costs, minimizes data transmission, and addresses security and privacy concerns critical for enterprise clients.

The inference speed of Om AI’s models reaches milliseconds, enabling real-time applications—from security systems to industrial monitoring and AIoT analytics. The company already collaborates with PC manufacturers like Apple, Lenovo, and HP, and its technologies are integrated into robots, drones, and other autonomous decision-making devices.

This year’s strategic priority is the development of the next-generation edge model VLX, which aims to improve video understanding and decision-making while reducing operational costs. As the industry shifts from cloud solutions to local edge systems, companies like Om AI are becoming key drivers of multimodal AI adoption in real-world scenarios.