step-1o-turbo-202506 is a multimodal reasoning language model developed by StepFun (阶跃星辰). It belongs to the Step-1o series, which is specifically designed to handle complex cognitive tasks such as advanced mathematics, coding, and logical deduction. The model utilizes a "System 2" thinking process, allowing it to perform internal chain-of-thought (CoT) reasoning to verify and refine its answers before outputting a final response.
As a "Turbo" variant, this version is optimized for a balance between reasoning depth and inference speed. It supports multimodal inputs, enabling it to interpret and reason over both text and images. This allows the model to solve complex visual problems, such as analyzing technical diagrams or performing spatial reasoning, with the same logical rigor applied to text-based queries.
The model has been benchmarked on competitive platforms such as the Chatbot Arena, where it has demonstrated strong performance in reasoning-heavy categories. It is part of StepFun's broader suite of large-scale models, which includes the Step-1 foundation models and the trillion-parameter Step-2 series, representing the company's focus on high-intelligence, multimodal AI solutions.