StepFun Launches StepEdge On-Device AI Models for Smartphones and Cars

StepFun has launched the StepEdge Edge-side Model Suite, comprising four products: StepEdge Base Model, StepEdgeAudio, StepEdgeGUI, and StepEdgeGen. Designed specifically for end devices such as smartphones and vehicles, this model series enables on-device Agents to respond rapidly while handling complex tasks through seamless edge-cloud collaboration. The StepEdge series features three core capabilities: First, ultra-low local latency—achieving edge-side tool call execution as fast as 0.1 seconds, allowing frequent, simple tasks to operate independently of the cloud. Second, full-modal privacy protection—multimodal data including text, vision, and audio is processed directly on the device, ensuring sensitive information never leaves the terminal. Third, a native edge-cloud协同 mechanism—simple tasks or those performed under poor or no network connectivity are handled on-device, while complex reasoning and long-chain tasks are offloaded to the cloud, balancing responsiveness, functionality, and cost efficiency. Additionally, the StepEdge series integrates with the StepInference NPU engine, which optimizes inference specifically for terminal hardware, further reducing end-to-end latency across diverse input modalities.