StepFun Launches Step Edge Models for Fast On-Device AI

StepFun has launched the Step Edge model family, including four models for basic tasks, audio, GUI, and generation. These models are designed for devices like phones and cars, aiming to move AI from the cloud to your device for faster responses and better privacy.

Key features include ultra-low latency (as fast as 0.1 seconds for local tool calls), full-modal privacy protection (text, images, and voice stay on your device), and native cloud-device teamwork—simple tasks are handled locally, while complex ones go to the cloud.

The models work with the Step Inference NPU engine to optimize performance on device hardware, reducing delays even further. This could make on-device AI agents more practical for everyday use in phones and cars.