TMTPOST — Chinese artificial intelligence developer Om AI announced on Tuesday the release of VLX, the world's first edge-side streaming multimodal model series designed specifically for interaction with the physical world. The new product suite integrates three distinct specialized architectural components comprising VLX-Flow for continuous visual tracking via incremental encoding, VLX-Seek for precise spatial localization through candidate region selection, and VLX-Go for short-horizon waypoint execution. Designed completely around lightweight parameter configurations ranging from 0.6 billion to 10 billion specifications, the unified software framework shifts model reasoning away from cloud-dependent request queues to run complex vision-language-action operations locally on edge hardware. Deployment vectors for lightweight, low-latency intelligence are expanding rapidly across the Chinese mainland as software ecosystems move toward decentralized hardware infrastructure like industrial drones, smart vehicles, and collaborative robotics. Rather than relying on standard frame-by-frame cloud processing, this localized approach allows hardware manufacturers to secure sub-second reaction loops while keeping proprietary operational data safely contained within local enterprise networks.