Amazon Nova 2.0 Pro Preview (medium) is a multimodal foundation model developed by Amazon, introduced as part of the Nova 2 generation in December 2025. It is designed for highly complex reasoning tasks, offering advanced capabilities in text, image, and video processing. The "medium" designation refers to a specific configuration of the model's extended thinking feature, which allows users to select a balanced intensity level between reasoning depth and execution speed.
The model features a 1-million-token context window, enabling the analysis of massive datasets, long documents, and lengthy video files within a single prompt. It is optimized for multi-step reasoning, long-range planning, and sophisticated agentic workflows, such as software engineering and multi-document synthesis. Compared to its predecessor, it demonstrates improved performance in perception tasks, including OCR and object detection.
In addition to its native multimodal understanding, the model supports built-in tools such as web grounding and a code interpreter to ensure responses are factually anchored. It is frequently utilized as a teacher model for knowledge distillation, allowing organizations to transfer its capabilities into more cost-effective, task-specific student models.