Amazon Launches Nova AI Models: Advanced Multimodal Foundation Models

Amazon has announced the launch of its next-generation foundation models, dubbed Amazon Nova, designed to revolutionize generative AI capabilities with industry-leading cost efficiency and performance. Integrated into the Amazon Bedrock platform, these models aim to streamline AI-powered applications for shoppers, sellers, advertisers, and enterprises.

What is Amazon Nova?

Amazon Nova represents a suite of advanced foundation models that support multimodal input—including text, images, and video—enabling diverse applications, such as understanding complex media and generating high-quality multimedia content.

The Nova family consists of several models tailored to specific needs:

• Amazon Nova Micro: A text-only model focused on ultra-low latency and affordability.

• Amazon Nova Lite: A cost-effective multimodal model optimized for speed across image, video, and text inputs.

• Amazon Nova Pro: A high-performance multimodal model balancing accuracy, speed, and cost.

• Amazon Nova Premier: Amazon’s most advanced multimodal model, designed for complex reasoning tasks and for use as the best teacher for custom model distillation (available in Q1 2025).

• Amazon Nova Canvas: A state-of-the-art model for generating high-quality images.

• Amazon Nova Reel: A leading-edge video generation model for creating dynamic multimedia content.

Innovations in Customization and Efficiency

Amazon Nova models stand out for their affordability, offering at least 75% cost savings compared to similar models within Amazon Bedrock. They also excel in customization, enabling users to fine-tune models with their proprietary data for tailored accuracy. Key features include:

• Fine-Tuning: Boost accuracy by training models with labeled proprietary data. The Amazon Nova model learns what matters most to the customer by analyzing proprietary data, which can include text, images, and videos. Amazon Bedrock then uses this data to create a private fine-tuned model optimized for the customer’s specific needs. This tailored model provides highly accurate responses, ensuring the AI aligns with the customer’s unique objectives and integrates seamlessly into their workflows.

• Distillation: Transfer specialized knowledge from advanced models to smaller, faster, cost-efficient models.

• Retrieval Augmented Generation (RAG): Ground responses in a user’s proprietary knowledge base for enhanced reliability. RAG enhances a model's responses by retrieving relevant information from external knowledge bases or datasets at runtime. Instead of training the model on all potential information, it retrieves data as needed to improve accuracy and grounding.

These features make Nova models suitable for agentic applications, where AI interacts with proprietary systems to execute multistep tasks efficiently.

Real-World Applications

Amazon Nova is already transforming industries, particularly in advertising and content generation:

• Amazon Nova Reel has been used to create imaginative video ads, such as the whimsical "Pasta City," where buildings are made of pasta and streets flow with marinara sauce.

• Amazon Nova Pro excels in video understanding, capable of analyzing and describing complex scenes, like a silent football game, with remarkable detail. The results provide detailed insights into the game's setting, the teams' uniforms, the actions of the players, and the conclusion of the play.

Looking ahead, Amazon plans to expand Nova with two groundbreaking models in 2025:

• A speech-to-speech model for humanlike verbal interactions including tone and cadence.

• An any-to-any modality model capable of processing and generating text, images, audio, and video seamlessly to simplify the development of applications.

Responsible AI and Safety

Amazon has integrated safety measures into all Nova models, including AWS AI Service Cards that provide transparency around use cases, limitations, and responsible AI practices. Learn more about those here. 

You can get started with Amazon Nova here. 

What This Means

Amazon Nova underscores the company’s ambition to lead in generative AI innovation. By combining multimodal capabilities, affordability, and robust customization, these models are poised to transform how businesses and individuals interact with AI-powered tools, especially for advertising products.

For Amazon, Nova represents not just a leap in AI technology but also a step toward democratizing advanced AI solutions, ensuring they deliver real-world value across industries. As more models roll out, including speech-to-speech and any-to-any modality, Amazon’s AI ecosystem will only grow more versatile and impactful.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing, image, and idea-generation support from ChatGPT, an AI assistant. However, the final perspective and editorial choices are solely Alicia Shapiro’s. Special thanks to ChatGPT for assistance with research and editorial support in crafting this article.