At the 2025 re:Invent global conference, Amazon Web Services (AWS) unveiled a significant expansion of its Nova product range, introducing four new Nova 2 models alongside a suite of additional services. The Nova 2 series comprises the Nova 2 Lite, Nova 2 Pro, Nova 2 Sonic, and Nova 2 Omni, all engineered to handle multimodal tasks encompassing text, image, video, and voice processing. Specifically, the Nova 2 Lite is a budget-friendly inference model tailored for everyday tasks; the Nova 2 Pro is a sophisticated reasoning agent built for complex endeavors, including programming; the Nova 2 Sonic is a voice-to-voice model optimized for conversational AI applications; and the Nova 2 Omni is a multimodal reasoning and generation model adept at processing diverse inputs and producing both textual and visual outputs. Moreover, AWS introduced Nova Forge, a customizable model service tailored for enterprise use. For an annual subscription fee of approximately RMB 708,000, customers can train exclusive model iterations using their proprietary data, thereby mitigating the issue of knowledge erosion during post-training model refinements.