Las Vegas — At the AWS re:Invent conference, Amazon Web Services (AWS) unveiled a new foundation model for generative AI (GenAI), which is meant to enable developers make use of the technology’s full potential while delivering desired results.
Andy Jassy, CEO of Amazon, introduced the new suite of GenAI tools powered by its advanced foundation models, Amazon Nova. Designed to simplify tasks across various industries, these models enhance capabilities in understanding, generating, and customizing content.
Amazon Nova models have been designed to support diverse tasks, including creating multimedia content and automating video understanding.
A new era of foundation models
The Amazon Nova lineup includes several models designed for different needs. Amazon Nova Micro is focused on text processing and stands out for its speed and cost-efficiency, making it ideal for tasks requiring quick and affordable solutions. Nova Lite builds on this by offering support for text, images, and videos while maintaining low latency and affordability. Nova Pro strikes a balance between accuracy, speed, and cost, making it suitable for general-purpose tasks.
The most advanced option, Nova Premier, is designed for complex reasoning tasks and excels in teaching smaller, customized models to perform efficiently. This model is set to be available in early 2025. Nova Canvas specializes in generating high-quality images, while Nova Reel transforms a single image into a dynamic video, enabling creative content production for businesses.
Practical applications
Amazon Nova models have been designed to support diverse tasks, including creating multimedia content and automating video understanding. For instance, Amazon Nova Pro demonstrated its ability to interpret a silent football game video, offering detailed insights into the scene, player actions, and game outcome.
Creative tools like Nova Canvas and Nova Reel are making waves in advertising. Brands using these tools have expanded their advertising reach, showcasing five times more products and doubling the creative assets per product. This innovation allows sellers to explore video advertising and test new marketing strategies, such as keyword-based creative content.
Simplified AI adoption
Using Amazon Nova through Amazon Bedrock, customers can access models from Amazon and other leading AI providers via a single API. This streamlined approach helps businesses experiment with different models to determine the best fit for their needs.
The models also excel in Retrieval Augmented Generation (RAG), ensuring accurate responses grounded in a company’s proprietary data. Additionally, agentic capabilities allow the models to interact with various APIs to complete multi-step tasks.
Cost-effective
Amazon Nova emphasizes cost-effectiveness. The Micro, Lite, and Pro models are priced at least 75% lower than similar options, making them an affordable choice for enterprises exploring AI-driven solutions. Their high-speed performance further enhances accessibility, supporting over 200 languages across various industries.
By addressing challenges like latency, customization, and cost, Amazon Nova represents a significant step in Amazon’s AI journey. Through applications ranging from simplifying shopping experiences to transforming advertising campaigns, Amazon continues to demonstrate the transformative potential of AI in everyday scenarios.
Amazon Nova Micro is a text-only model designed for low-latency, cost-effective operations. Nova Lite processes text, images, and videos with impressive speed and affordability. Nova Pro provides a balanced solution, combining accuracy, speed, and cost-efficiency. Nova Premier, arriving in early 2025, tackles complex reasoning tasks and supports the development of smaller, efficient models. Nova Canvas delivers advanced image generation, while Nova Reel brings still images to life by converting them into short videos.
These tools offer scalable, flexible AI solutions to drive innovation and adoption across industries.
Updated Dec. 4, 2024