Search
StepOn AI has just declared the formal rollout of its latest model, Step 3.5 Flash 2603, accessible to all users enrolled in the Step Plan. This model represents an upgraded iteration of Step 3.5 Flash, preserving its hallmark traits of swift response times and cost-effectiveness. Simultaneously, it introduces a low think mode designed to minimize token usage in specific scenarios and bolster output efficiency. The model has also benefited from refined training tailored for Coding and Agent frameworks, thereby elevating the overall user experience, stability, and token utilization efficiency. In its default high mode, the model maintains peak inference performance, achieving a 14% decrease in token consumption. Conversely, in low mode, token consumption plummets by 56%, resulting in a substantial enhancement in inference efficiency.
-
C114 Communication Network -
Communication Home
