Skip to main content

Step Astronomy Launches Edge AI Models for Phones and Cars

Step Astronomy has officially launched the Step Edge Series Terminal Models, a new line of AI models tailored for mobile phones, in-vehicle systems, and other edge devices. The series includes four versions: Basic, Audio, GUI, and Gen. These models are designed to process text, images, and audio locally, enabling features such as screen understanding, speech recognition, interface operations, and even image generation—all without needing a constant cloud connection.

According to the company, the Step Edge series achieves a call delay as low as 0.1 seconds. For simple, frequent tasks or in weak network environments, everything can be handled right on the device. More complex reasoning tasks are offloaded to the cloud. This hybrid approach significantly cuts down on network dependency and cloud computing costs. Plus, sensitive data never has to leave the device, which is a big win for privacy and security.

Image

To further boost performance on edge hardware, Step Astronomy also introduced its self-developed Step Inference NPU inference engine. This engine is designed to dramatically reduce inference latency on terminal chips, making real-time AI applications more practical.

The company claims that the Step Edge series ranks first in 29 core evaluation metrics, showcasing leading multimodal capabilities on the terminal side. This move marks a significant step in the push to bring AI directly to end-user devices, promising more responsive and private experiences for smartphones, cars, and other smart gadgets.

Key Points

  • Step Edge series includes four models: Basic, Audio, GUI, and Gen.
  • Processes text, images, and audio locally with a 0.1-second response time.
  • Hybrid architecture: local for simple tasks, cloud for complex reasoning.
  • Self-developed NPU inference engine reduces latency on terminal chips.
  • Claims top rankings in 29 evaluation metrics for terminal-side multimodal AI.