At its ongoing re:Invent conference, Amazon Web Services (AWS) unveiled the Nova family of artificial intelligence (AI) models on Tuesday. The new suite features five large language models (LLMs), three of which are dedicated solely to text generation. In addition, Nova includes models for image and video generation. According to the company, these models are designed to offer enhanced intelligence and competitive pricing, and they can be accessed through Amazon Bedrock.
AWS Introduces Nova AI Models
Amazon provided further details about the Nova models in a recent post, showcasing the five different LLMs now available. Company CEO Andy Jassy mentioned that a sixth model, named Nova Premier, is anticipated to launch in 2025.
The trio of text-only models—Nova Micro, Nova Lite, and Nova Pro—offers varied capabilities. Nova Micro processes only text input and is noted for its low latency, boasting a context window of 128,000 tokens.
Meanwhile, Nova Lite can accept images, videos, and text as input while still generating only text. Nova Pro stands out as the most versatile multimodal AI model among the three, enabling a broader range of tasks. Both Nova Lite and Nova Pro feature a context window of 300,000 tokens.
Additionally, the series includes two “creative content generation models.” Nova Canvas is an image generation model that takes both text and image inputs, being positioned as a resource for advertising, marketing, and entertainment sectors.
Lastly, Nova Reel is a video generation model capable of creating short videos from text and image prompts. It also provides users with the ability to manipulate camera motion through natural language commands. These innovative models are accessible to enterprise clients via the Amazon Bedrock platform.
