Overview Our Story Customers Partners

Amazon Nova Understanding Models

Frontier intelligence and industry leading price performance

Amazon Nova Understanding Models

Amazon Nova understanding models are state-of-the-art models that deliver frontier intelligence and industry-leading price performance.

Amazon Nova Micro is a text-only model that delivers the lowest latency responses at a very low cost.
Amazon Nova Lite is a very low-cost multimodal model that is lightning fast for processing image, video, and text inputs to generate text outputs.
Amazon Nova Pro is a highly capable multimodal model with the best combination of accuracy, speed, and cost for a wide range of tasks.
Amazon Nova Premier is our most capable model for complex tasks and teacher for model distillation.

Amazon Nova understanding models support over 200 languages, text and vision fine-tuning, and easy integration with proprietary data and applications through Amazon Bedrock features like Bedrock Knowledge Bases and Bedrock Agents. Using the appropriate model and techniques like distillation, you can get the optimal intelligence, latency, and costs for your needs with Amazon Nova understanding models. You can learn more about Amazon Nova understanding models in the Amazon Nova technical report and the Amazon Nova Premier technical report.

Key capabilities

Frontier intelligence

Amazon Nova understanding models deliver state-of-the-art text and visual intelligence, with native support for plain text, documents, image, and video understanding.

Customization and Knowledge Distillation

Amazon Nova understanding models support text and vision fine-tuning, with text, image, and video inputs as reference data for training. Amazon Nova models can also be used with Amazon Bedrock Model Distillation to create highly capable, cost-effective, and low latency versions of Nova Pro, Lite, and Micro, for your specific needs.

Fast, cost-effective models

Amazon Nova understanding models are among the fastest and most cost-effective models in their respective intelligence classes. With new latency-optimized inference options, enterprises can achieve even faster response times for their mission-critical AI applications without compromising quality.

Retrieval Augmented Generation (RAG) and enterprise data grounding

Amazon Nova understanding models excel at in-context learning (ICL) and grounding responses based on contextual data provided in approaches like RAG. This is seen in Amazon Nova model scores in the Comprehensive RAG Benchmark (CRAG) evaluation. Customers can easily integrate Amazon Nova understanding models with their proprietary data and applications using Amazon Bedrock features like Bedrock Knowledge Bases.

Function calling and agentic workflows

New gen AI applications are increasingly using AI agents to interact with external services and APIs, and to perform tasks in the customer’s environment. Amazon Nova models excel at agentic workflows and are the state-of-the-art on on benchmarks that evaluate this capability, like Berkeley Function Calling Leaderboard (BFCL), VisualWebBench, and Mind2Web.

Introducing Amazon Nova Premier