Step Star has released three models at WAIC, focusing on multimodal capabilities.
Step-2: A trillion-parameter MoE model, which currently requires an application to experience and is approved quite quickly. However, the open platform is too rudimentary and lacks a playground, so one can only experience it by calling it themselves.
Step-1.5V: A hundred-billion-parameter multimodal model, which not only enhances image understanding capabilities but also supports video understanding.
Step-1X: An image generation model, with the DiT architecture, available in three different parameter sizes: 600M, 2B, and 8B. It has been optimized for Chinese culture and elements.