Amazon Nova Pro is a multimodal foundation model developed by Amazon, designed to provide a balance of intelligence, speed, and cost-efficiency for complex reasoning tasks. It is part of the first generation of the Amazon Nova model family, positioned as a high-performance, mid-range model capable of processing diverse input types, including text, images, and video.
The model features a 300,000-token context window, which allows it to ingest and analyze extensive document collections or long-form video content. It is optimized for high-throughput processing, making it suitable for enterprise-grade applications such as automated document analysis, data extraction from complex layouts, and visual question answering. Nova Pro was trained on a massive multimodal dataset covering more than 200 languages.
In addition to its core language and visual understanding, Nova Pro supports advanced agentic workflows and tool-use capabilities. It is designed to act as a "teacher" model for model distillation, enabling the creation of smaller, task-specific models that maintain high accuracy while reducing operational costs. Its broad reasoning capabilities also extend to software development, including code generation and multi-step problem solving.