In the current phase where competition in large models is shifting from parameter scale to real-world deployment capability, a group of Chinese companies focused on edge AI is gaining attention, and Om AI Technology is one of them.
Founded in 2021, the company has chosen not to pursue extremely large cloud-based models, but instead focuses on edge-side general-purpose multimodal vision models, aiming to bring AI into real devices such as PCs, cameras, and robots.
At the BEYOND Expo 2026 media day, Om AI Technology showcased its AI-native content creation product developed in collaboration with Lenovo, OttoBox AI Studio. Designed for media professionals and content creators, it leverages local AI computing power to provide capabilities such as video analysis, asset matching, script generation, and rapid video production.
The company positions it as a content creation companion for the AI-native era, aiming to improve creative efficiency.
Compared with many AI companies that move from general-purpose models into application layers, Om AI starts from a more industry-driven foundation. The team has long been deeply involved in the media and audiovisual industry, and therefore emphasizes building models based on real-world problems rather than looking for nails with a hammer.
Dr. Zhao Tiancheng, CEO of Om AI, noted that long-term industry experience not only helps the team deploy models faster, but also provides access to large amounts of high-quality real-world data. In their view, true multimodal capability is not just about recognizing images and text, but about understanding video, audio, and text simultaneously.
One of the company’s key technical focuses is video understanding under low-parameter models. Compared with traditional approaches that rely on extremely large parameter counts and cloud-based GPU resources, Om AI emphasizes a small, precise, and fast edge-model approach.
By reducing model size, AI can run directly on local devices, lowering inference costs and reducing data upload requirements, while also addressing enterprise concerns around data security and privacy.
This edge deployment advantage is particularly significant in large-scale video analysis scenarios. The company states that its models can achieve millisecond-level inference speed, making them suitable for real-time applications such as security, industrial inspection, and AIoT analytics.
Currently, Om AI’s AI business spans three major areas: AI PCs, AIoT, and embodied intelligence. In addition to collaborations with Apple, Lenovo, and HP, its models are also applied to robots, robotic dogs, and drones, enabling these devices to gain autonomous decision-making and action capabilities.
Om AI is also exploring inclusive AI applications. For example, its Haoma App, designed for visually impaired users, enables object search and assisted navigation through smartphones or AI glasses.
This year, the company’s key strategic priority is the launch of its next-generation edge multimodal model VLX, which aims to further improve video understanding and decision-making while continuously reducing operational costs.
As the AI industry shifts from cloud-based competition toward on-device deployment, companies like Om AI are becoming key drivers of real-world multimodal AI adoption.
0 Commentaires