Om AI, the Chinese startup also known as 联汇科技 (Lianhui Technology), has secured hundreds of millions...

Om AI, the Chinese startup also known as 联汇科技 (Lianhui Technology), has secured hundreds of millions of yuan in new funding led by Qianhai Fuhai Fund, alongside the open-sourcing of its VLX-Seek 1.5, an edge-native streaming multimodal model series with 3B and 10B parameter versions. The funding and release underscore a strategic bet on edge-native AI architectures for physical-world applications like robotics and drones. The company claims its 3B model outperforms NVIDIA's LocateAnything-3B on key benchmarks, including object detection, referring expression comprehension, and drone-view perception tasks.
This move positions Om AI at the center of a rising trend in physical AI—edge-native processing. Unlike cloud-dependent models that rely on remote servers, VLX-See is designed from the ground up for edge devices, optimizing for latency, power, and deployment cost. This is a fundamental architecture shift, akin to the transition from cloud computing to cloud-native. By open-sourcing VLX-Seek 1.5, Om AI aims to establish its architecture as a standard, attracting developers and building an ecosystem that could rival the cloud-centric approaches of NVIDIA, Google, and Tesla.

For builders and investors, the implications are concrete: the race for physical AI is no longer just about model size but about unit parameter value on edge hardware. Om AI's benchmarks suggest that small, edge-native models can compete with larger cloud-based rivals in specific use cases, potentially enabling faster, more autonomous deployments. Investors should watch whether this edge-native approach gains traction beyond Om AI, potentially reshaping the compute infrastructure for robotics and IoT. Builders exploring edge AI for physical-world applications should evaluate VLX-Seek 1.5 as a viable alternative to cloud-dependent solutions. #EdgeAI #PhysicalAI #Robotics #VLX #OmAI #AIInfrastructure