Nova Premier is the largest and most capable multimodal foundation model in Amazon's Nova family. It is designed to handle complex reasoning tasks, deep contextual understanding, and multistep planning. The model is positioned as the flagship of the Nova lineup, offering the highest level of intelligence for sophisticated enterprise workflows.
As a multimodal model, Nova Premier can process inputs across text, images, and video. It features an expansive context window of 1,000,000 tokens, enabling the analysis of massive datasets such as large codebases, extensive document collections, and long-form video content in a single prompt. This capacity makes it suitable for advanced information retrieval and high-fidelity data synthesis.
Beyond direct inference, Nova Premier serves as a primary "teacher" model for model distillation. This allows developers to use its outputs to train smaller, more efficient versions of other Nova models (such as Pro, Lite, or Micro) for specific tasks. This architectural role facilitates the creation of specialized models that retain high accuracy while operating at significantly lower latency and cost for production deployments.