Step3 is a multimodal reasoner that handles both text and images with a methodical, step-by-step approach — true to its name. It tends to break down complex problems into traceable chains of thought, making its reasoning process relatively transparent. As an open-weight model under Apache 2.0, it trades some raw capability ceiling for accessibility and deployability.