Amazon Takes On Microsoft and OpenAI with Nova

At the ongoing annual AWS re:Invent conference in Las Vegas, the e-commerce giant launched Amazon Nova, its in-house family of foundational models, available exclusively through Amazon Bedrock.

This new generation of models is built to support a variety of generative AI tasks, offering businesses an innovative way to lower costs and reduce latency while addressing complex needs.

“They’re really cost-effective. They’re about 75% less expensive than the other leading models in Bedrock," said Amazon chief Andy Jassy.

He explained the innovative features of the new Nova models, stressing their affordability and low latency. “The models are integrated with the distillation feature, which allows you to infuse the intelligence of larger models into smaller models that are more cost-effective and offer lower latency,” he said.

Jassy further added that the Nova models are deeply integrated with Bedrock knowledge bases, enabling users to “use RAG to ground your answers in your own data.”

Amazon Nova provides both understanding models, which process text, image, or video inputs to generate text outputs, and creative content generation models that can create images and videos from text or images.

The understanding models include Amazon Nova Micro, Lite, Pro, and Premier. Each model is optimised for specific tasks, ranging from text summarisation and translation to complex document processing and multimodal interactions. 

For instance, Amazon Nova Micro is a text-only model that delivers low-latency responses, making it ideal for text-based tasks such as content classification or interactive brainstorming. 

On the other hand, Amazon Nova Pro supports high-level tasks such as visual question answering and video understanding, making it perfect for enterprise workloads, including financial document analysis and sophisticated AI agent tasks. The model is on par with Anthropic Claude Sonnet 3.5, GPT-4o, and Gemini 1.5 Pro.

In terms of creative content generation, Amazon Nova introduces two powerful models, Nova Canvas and Nova Reel. 

Canvas allows users to generate studio-quality images with editing capabilities, including inpainting and background removal, while Reel enables the creation of professional-quality videos from text prompts and images. These models are highly adaptable, making them suitable for use in marketing, advertising, and entertainment industries.

The flexibility of Amazon Nova lies in its ability to be customised. Businesses can fine-tune the models to understand specific industry terminology, ensuring that the models align with brand voice and requirements. This adaptability is particularly beneficial for industries such as legal or healthcare, where precision in language and understanding is critical.

With robust safety controls, including watermarking for creative content, and the ability to handle multimodal inputs and generate content across various media types, Amazon Nova is a versatile and scalable solution for businesses. It is currently available in select AWS regions with a pay-as-you-go pricing model.